Predicting Plasticity and Constraint Under Climate Change

Oecologia (2011) 165:237–248 DOI 10.1007/s00442-010-1789-8

GLOBAL CHANGE ECOLOGY

Environmental controls on the phenology of moths: predicting plasticity and constraint under climate change

Anu Valtonen • Matthew P. Ayres • Heikki Roininen • Juha Po¨yry • Reima Leinonen

Received: 26 April 2010 / Accepted: 25 August 2010 / Published online: 30 September 2010 Springer-Verlag 2010

Abstract Ecological systems have naturally high inter- should be most immediately responsive in phenology to annual variance in phenology. Component species have climate warming, but variably so depending upon the presumably evolved to maintain appropriate phenologies minimum temperature at which appreciable development under historical climates, but cases of inappropriate phe- occurs and the thermal responsiveness of development rate. nology can be expected with climate change. Understand- Photoperiodic modification of thermal controls constrains ing controls on phenology permits predictions of ecological phenotypic responses in phenologies to climate change, but responses to climate change. We studied phenological can evolve to permit local adaptation. Our results suggest control systems in Lepidoptera by analyzing flight times that climate change will alter the phenological structure of recorded at a network of sites in Finland. We evaluated the the Finnish Lepidoptera community in ways that are pre- strength and form of controls from temperature and pho- dictable with knowledge of the proximate physiological toperiod, and tested for geographic variation within spe- controls. Understanding how phenological controls in cies. Temperature controls on phenology were evident in Lepidoptera compare to that of their host plants and ene- 51% of 112 study species and for a third of those thermal mies could permit general inferences regarding climatic controls appear to be modified by photoperiodic cues. For effects on mid- to high-latitude ecosystems. 24% of the total, photoperiod by itself emerged as the most likely control system. Species with thermal control alone Keywords Lepidoptera Light-trap Photoperiod Temperature Thermal sum

Communicated by Je´rome Casas. Introduction Electronic supplementary material The online version of this article (doi:10.1007/s00442-010-1789-8) contains supplementary material, which is available to authorized users. Ecological systems have naturally high interannual variance in phenology, i.e., in the seasonal timing of periodic A. Valtonen (&) H. Roininen life cycle events (Rathcke and Lacey 1985). These events Department of Biology, University of Eastern Finland, are strongly influenced by interannual variation in envi- P.O. Box 111, 80101 Joensuu, Finland e-mail: anu.valtonen@uef.fi ronmental conditions, especially temperatures. Populations are generally well adapted in their phenological adjust- M. P. Ayres ments to interannual climatic variation. The eggs of many Department of Biological Sciences, Dartmouth College, herbivorous insect species tend to hatch when their host Hanover, NH 03755, USA plants leaf out (van Asch and Visser 2007), insectivorous J. Poÿry birds tend to breed as their prey become abundant (Perrins Research Programme of Biodiversity, and McCleery 1989), insects enter and break diapause as Finnish Environment Institute, 00251 Helsinki, Finland appropriate for the progression of season (Denlinger 2002), R. Leinonen and adult moths typically emerge at the appropriate time Rauhalantie 14 D 12, 87830 Nakertaja, Finland for oviposition (e.g., Tammaru et al. 2001). 123 238 Oecologia (2011) 165:237–248

Plastic responses to environmental cues can help indi- meaningful development occurs (developmental threshold) viduals synchronize their life cycle to interannual variation and of the slope of development rate versus temperature in climate and resources. Phenotypic plasticity is the pro- above the threshold (Trudgill et al. 2005). The phenology duction of alternative phenotypes by the same genotype of species whose development is temperature driven will under different environmental conditions (Nylin and be predictably responsive to interannual temperature vari- Gotthard 1998), for example changes in the growth rate and ation, including directional climate change. However, therefore timing of reproduction depending upon temper- species within a community could vary in their respon- ature. Natural selection on phenotypic plasticity can siveness due to differences in their developmental thresh- enhance the fit between an organism and its environment. olds, differences in their temperature sensitivity when However, in some cases, climate change is already above the threshold, and differences in the extent to which exceeding the range of climatic conditions under which photoperiod modifies the thermal response. adaptive phenological plasticity has evolved (Post and Photoperiod is another well-known modifier of insect Forchhammer 2008), and there are increasing examples of development and therefore phenology since it frequently phenological disruptions of populations, communities, and provides the primary signal for initiation of diapause ecosystems (Visser and Holleman 2001; Willis et al. 2008; (Bradshaw and Holzapfel 2010). Photoperiodically enforced Post et al. 2009). A challenge is to understand and predict diapause in fall is presumed to be adaptive by preventing which species are most susceptible to disruption of inappropriate development in the late fall or winter, even appropriate phenological patterns and which species will when temperatures are permitting (Danilevskii 1965; tend to be most robust in their capacity for adaptive Tauber et al. 1986; Leather et al. 1993). Some insect species plasticity. require a critical photoperiod to terminate diapause in spring Lepidoptera are a good model system to study the nature or early summer (Tauber et al. 1986; Danks 1987; Leather of phenological plasticity because they are diverse, abun- et al. 1993). Some species also display a period of devel- dant, reasonably well understood with respect to their opmental inactivity during summer (aestivation), the ter- physiological ecology, and a globally important component mination of which can likewise be subject to photoperiodic of primary consumers in terrestrial ecosystems (Stamp and control (Masaki 1980; Tauber et al. 1986). Irrespective of Casey 1993; Scoble 1995). During the past decades, climatic variation, complete photoperiodic control leads to Lepidoptera have experienced both advanced phenology invariant interannual phenology with respect to Julian date. (Kuchlein and Ellis 1997; Forister and Shapiro 2003) and Temperature and photoperiod can be co-determinants of increased voltinism (Altermatt 2010). We studied the phenology. For example, temperature can be the dominant diverse community of Finnish nocturnal moths to evaluate driver of development rate and therefore phenology after patterns in the prevalence of four general theoretical pos- the critical photoperiod to terminate diapause is reached. sibilities for environmental controls on phenology: (1) However, many diapausing species seem to exhibit a temperature, (2) Photoperiod, (3) temperature and photo- gradual loss during late autumn and early winter in sensi- period together, and (4) regional adaptation of populations tivity to factors that maintain diapause (Tauber et al. 1986; in the form of their responses to temperature and/or Danks 1987), in which case development can proceed in photoperiod. spring as soon as temperatures rise above the develop- Temperature is frequently a dominant driver of pheno- mental threshold. logical state in insects because their development rate is Finally, there can be divergence among populations in strongly temperature-dependent (Gillooly et al. 2002). The different regions in the form of their responses to temper- generalized function, which is explicable in terms of ature and/or photoperiod. This is the least well studied of thermodynamics and enzyme function, is for development our general models of phenological controls, but geo- rate to be imperceptibly low at low temperatures, increase graphical variation in insect seasonal cycles has been with increasing temperatures across a range that is typically recorded across different latitudes, altitudes, and proximity broad with respect to realized temperatures during the to the center of large land-masses (Danilevskii 1965; growing season, and then decelerate and decline at tem- Leather et al. 1993) and can involve both continuously peratures that are warm with respect to that to which the varying characteristics (for example, diapause-inducing population is adapted (Davidson 1944; Logan et al. 1976; photoperiods or diapause duration) and disjunctly varying Sharpe and DeMichele 1977; van der Have 2002). Since traits (e.g., voltinism) (Tauber et al. 1986). Evidence for the time of Reáumur (Egerton 2006), it has been recog- population differentiation in the controls on phenology nized that a linear approximation of this relationship can implies a capacity for adaptive adjustments to climate explain much of the interannual variation in phenological change. events. At their simplest, such thermal sum models only We exploited historical records from across Finland of require an estimate of the minimum temperature at which the flight phenology of 112 species of moths to test the 123 Oecologia (2011) 165:237–248 239 applicability of these alternative models of controls on the phenology of high-latitude Lepidoptera. We asked what proportion of species have phenologies strongly controlled by temperature compared to species for which the timing of

flight is controlled or modified by photoperiod, and we TapTrap Weather station evaluated whether lability in phenological controls is rare Finland or common. We asked what kind of species tend to follow 0 100 200 km which system of control: Lepidoptera could be quite uniform in the controls on their phenology or there could be variation among clades (e.g., Geometroidea vs Noctuoi- dea), ecologically defined groups (species with different season of flight, overwintering stage or diet breadth), and/ or geographic distribution. Results provide an enhanced understanding of plasticity and constraint in insect phenology and permit predictions of the effects of climate change on a globally important group of primary consumers, which promotes understanding of when and where trophic systems are likely to become destabilized by phe- 60 62 64 66 68 70 nological disruptions versus being as resilient to climate change as they are to historical norms of interannual cli- 18 20 22 24 26 28 30 32 matic variation. Fig. 1 Locations in Finland of the 16 moth trapping sites and the weather stations used in analyses Materials and methods calculated. This left us with 450 timeseries representing Moth data 112 species. For analyses, we scored the day of capture as the middle Our analyses were based on a 12-year (1993–2004) time date of the weekly trapping period and converted this to series of moth captures (light traps) in 16 forested sites in days from the last winter solstice. For records from before Finland (Fig. 1; Finnish Moth Monitoring Scheme Noct- 1998, the date of emptying traps was not uniquely recorded urna, Finnish Environment Institute). Captured individuals for each trap but was rounded off to Saturday of the trap- were recorded with approximately weekly accuracy ping week. For these years, we scored the day of capture as throughout the flight period of macro-moths (April–May Thursday assuming that operators generally followed the through September–November). Data include species guideline of emptying traps each Sunday (Environment belonging to superfamilies Lasiocampoidea, Bombycoidea, Data Centre 1994). Geometroidea, Noctuoidea and Hepialoidea. Systematics and nomenclature follow Kullberg et al. (2009). Temperature data and thermal sum calculations Different generations can be impossible to separate and therefore we excluded 194 of 711 species of the original To calculate thermal sums we obtained daily minimum and data that have more than one generation per year and 5 maximum temperatures from 35 stations of the Finnish other species that fly in both fall and spring because they Meteorological Institute (Fig. 1). In some cases, the data overwinter as adults. We analyzed ten sites with contin- from one station did not cover the entire study period and uous trapping through every season of all 12 years and an therefore several different weather stations had to be additional six sites with data from each year, but with selected in the vicinity of each trap. The closest weather some missing dates. From these six sites, only time-series station with data was used when it was within 30 km dis- of species flying during the recorded months were selec- tance and 50 m altitude of the trap. Otherwise, we esti- ted. Then, we dropped species that were too rare for mated the daily maximum and minimum temperatures analysis (\120 total captures over 12 years) and excluded using the nearest records adjusted for temperature lapse years with less than 10 captures, leaving us with 494 rates (Online Resource 1). timeseries (each representing one species at one site) For each day, degree-hours were calculated as the esti- representing 154 species. Finally, we dropped species that mated hourly temperature minus the base temperature, did not meet these criteria for at least ten trap-year summed for the day, and divided by 24 to yield degree- combinations, from which the timing of peak flight is days. Thermal sums for each day of each trap were then 123 240 Oecologia (2011) 165:237–248 calculated as the accumulated sum of the degree-days to The base model against which the above were compared date. We estimated hourly temperatures from daily maxi- was solar day (Sday): the average day of peak flight (across mum and minimum temperatures by applying two sine- years and sites) counted from last winter solstice. In this functions and a square-root function (Cesaraccio et al. sense, Sday served as a null model. However, it also rep- 2001; details in Online Resource 2). resents a theoretical possibility. Species could be selected Thermal sum models for insects at the latitude of our to fly on about the same calendar date even when distrib- system have frequently assumed a base temperature uted across a range of thermal and photoperiodic envi- (developmental threshold) of 5C. However, base temper- ronments. In this case, Sday could provide a better fit than atures have been reported to vary from at least -5to models based on just temperature and/or photoperiod. ?10C among species in the same environment (Pritchard et al. 1996) and we wanted to allow for this theoretical Ecological characteristics of species possibility. Thus, we estimated the base temperature for each species that allowed the best goodness of fit from all We compared the phenological controls among (1) super- integers between -5 and ?10C. To parameterize the families, (2) season of flight, (3) overwintering stage, (4) model which allowed for effects of both temperature and diet breadth, (5) global distributional pattern, and (6) local photoperiod, we simultaneously solved for the base tem- distributional pattern. Classifications were based upon perature and the date on which the thermal sum model Finnish identification guides and related literature (list in started (representing the end of diapause and start of Online Resource 3). Flight seasons were classified as spring development) that maximized goodness of fit. For this, we (May or before), summer (June–August) or fall (September considered all start dates on which photoperiod first or after) based on average flight day across all years and reached 6, 7, 8, …, 18 h (which covers the variation of day sites. Overwintering stage was classified as adult, egg, small lengths experienced within the year and across the latitudes larva, full grown larva, larva (several stages or stage not of study). known), pupa or several (only one species). Diet breadth was classified as monophagous (only one host species), Models strongly oligophagous ([1 host species, but only one host genus), oligophagous ([1 genus but only one host family), Median flight dates were calculated for each species in or polyphagous ([1 host family). Global distributions were each year at each site as the mid-date of the weekly trap- classified as Holarctic, Palearctic (including Eurasian spe- ping period that contained the 50th percentile of cumula- cies and species distributed from Europe to SW Asia or from tive annual captures for that species in that year at that site. Europe to Siberia) or West Palearctic (including species We then compared the ability of the alternative theoretical with European distribution or from Europe to Black Sea). models to predict median flight date for each species Local distributions were classified as only South Finland, (across all years and sites). South to Central Finland (including regions of Pohjanmaa For the model based on temperature alone (Tsum), we and Kuusamo), or South to North Finland. solved for the best-fit model to predict median flight date across years and sites from only thermal sum. This Statistical analysis involved identifying the combination of developmental threshold and thermal sum that minimized the sums of We compared the ability of the alternative theoretical square error (SSE) for that species (with the accumulation models to predict the observed peak flight trough evalua- of thermal sums always starting at winter solstice). tion of the corrected Akaike Information Criteria (AICc, For the Photoperiod model, we solved for the best-fit Anderson 2007) and ranked the models accordingly. We model to predict median flight date based on a photoperi- also calculated root mean square errors (RMSE), which odic threshold (representing minimum or maximum described the accuracy of the estimate in days. For evalu- required day length for species with peak flight before or ations of Tsum and Tsum \ Photoperiod, we were also able after mid-summer, respectively). 2 For the model of joint control by temperature and to estimate the proportion of variance explained Radj photoperiod (Tsum \ Photoperiod), we extended the Tsum relative to solar day. For those species with restricted lat- model by also allowing for a photoperiodic threshold at itudinal distributions in our sample, photoperiod was nec- which thermal sums would begin to accumulate. For this, essarily redundant with Sday so we made no effort to we identified the best-fit model for each species from distinguish between them with model comparison. All among 208 different sets of thermal sums calculated with formulas are given in Online Resource 4. base varying between -5 and ?10C and the critical day We then tested for variation among sites, i.e., population length varying between 6 and 18 h of daylight. differentiation of those species with evidence for 123 Oecologia (2011) 165:237–248 241 phenological control by Tsum, Photoperiod or Tsum \ Pho- Table 1 Frequency (and % of total) of species with different top toperiod. Under this model, the timing of flight in all popu- models for control of flight phenology lations is controlled by the same mechanism, but the Top models for control Number of %of thresholds are different. We tested for this in two ways. First, of flight phenology moth species total we calculated a single metric of summer thermal conditions at Tsum (T) 26 23 each site (average annual total thermal sum of the site, Tsum \ Photoperiod (TP) 18 16 TotTsum, calculated from the entire study period with TorTP 13 12 base =5C, starting and ending at the winter solstice) and used Photoperiod (P) 6 5 that as an explanatory variable in a linear model to explain the Photoperiod or Solar day (P or S) 21 19 difference between observed and predicted day of peak flight. TotTsum was negatively correlated with latitude but also Solar day (S) 8 7 incorporated a tendency towards increasingly continental TorS 1 1 climate from west to east in Finland. We also evaluated an TP or S 6 5 ANOVA model with site as a fixed factor, representing S or T or P 2 2 populations (Pop). Details of the structure of these models are S or P or TP 3 3 given in Online Resource 4. In the last phase, we classified the S or T or TP 1 1 species based on the best available model (or models) using T or P or TP 1 1 differences in AICc, RMSE and number of sites in the S or P or T or TP 6 5 judgement (Online Resource 5). Pearson’s Chi-squared test Total 112 was used in contingency analyses to test for differences in the frequencies of best models among ecological groups. All analyses were conducted with the R program, version 2.7.1 (R could not themselves be distinguished (usually because the Development Core Team 2008). available data were from similar latitudes, which made Sday and Photoperiod statistically redundant). Thus, if photoperiod is regarded as a more parsimonious proximate mecha- Results nism than solar day, there were 27 of 112 species for which photoperiod by itself emerged as the prevalent control on Competing models of environmental controls flight phenology. on phenology Surprisingly to us, there were eight species for which solar day emerged as a superior model to alternatives The raw variation in timing of median flight within species involving photoperiod and/or temperature. Particularly was generally high, which provided a basis for comparing clear cases were Dysstroma citratum, Epirrita autumnata, alternative models. The root mean square error (RMSE) of and Eupithecia pusillata (Di C 20 compared to closest the null model (Sday) ranged from 3.4 to 22.3 days (1st competing models). We were conservative in assigning quartile, median, 3rd quartile being 6.7, 7.8, and 9.2 days, species to this category and only rejected the more parsi- respectively). Model comparisons (Online Resource 5) monious alternatives of thermal sum and/or photoperiod if produced informative results for 92 of 112 moth species Di C 3.5 and the RMSE was lower (Online Resource 5). (Online Resource 6). This left 20 species for which we were unable to support or For 26 species, thermal sum (Tsum) was a superior model rule out controls by Tsum, Sday, Photoperiod, and com- of phenological control relative to alternatives of Photope- binations thereof (Table 1). For six of these species, there riod, Tsum \ Photoperiod,orSday (Di C2; Table 1; Online was little variation, and therefore little information, across Resource 6). For another 18 species, Tsum \ Photoperiod the observed flight times (RMSE \7 day in Sday). How- emerged as the single best model (with five of these showing ever, this group also included 14 species with plenty of additional evidence of differential thresholds among popu- variation in flight times (six species with RMSE [10 day), lations), and for 13 more species, Tsum and Tsum \ Pho- including Thera cognata (RMSE = 22 days). Altogether, toperiod were superior to alternatives but not themselves the results imply that our ensemble of possible models was distinguishable with available information (Di \2). Thus, inadequate to capture the true controls on phenology for there was evidence for thermal control, sometimes modified about 18% of the studied species. by photoperiod, in 57 of 112 species. Photoperiod by itself Of 65 species with data across a range of latitudes (more emerged as the single best model for six species (with four than one area in Online Resource 6), there was evidence for showing population differentiation). There were another 21 geographic variation in phenological thresholds in 17 species where photoperiod or solar day were demonstrably species (26%). Altogether nine species (five species with better models than alternatives including temperature, but Tsum \ Photoperiod as their best model and four species 123 242 Oecologia (2011) 165:237–248 with photoperiodic control) displayed population differen- than 4 days which is about as good as possible given the tiation. Of our two candidate metrics, population (Pop) weekly sampling resolution. Not surprisingly, there was provided the most information and TotTsum was superior some tendency for better model fits to species that occurred to Pop only for one species. We also interpreted the eight at fewer sites, but there were a total of 31 species (14 that species where Sday was superior to models involving Tsum occurred at two or more areas) with RMSE \5. For those and/or photoperiod as cases that most probably reflect with Tsum by itself as the best model, the average RMSE population differentiation. was 5.2 days (range 3.0–8.2, n = 26), for Photoperiod For many species, our limited suite of models permitted models 6.3 days (range 3.3–13.0, n = 27) and for very satisfactory predictions of flight phenology. For Tsum \ Photoperiod models 5.7 days (range 4.0–8.8, example, for Agrochola litura, Cybosia mesomella, Eilema n = 18). Tsum models explained an average of 54% of the lutarellum, Eulithis pyraliata, Martania taeniata, Mythi- variation compared to solar day alone (range 33–78%) and mna ferrago, Odontosia sieversi, Photedes minima, and Tsum \ Photoperiod models explained an average of 43% Tholera decimalis, the root mean square errors were less of the variation (range 9–63%). Figure 2 shows graphical

f(solar day) f(thermal sum) f(photoperiod) photoperiod) Orthosia gothica 20−Apr 10−May 30−May

20−Apr 10−May 30−May 20−Apr 10−May 30−May 20−Apr 10−May 30−May 20−Apr 10−May 30−May Hydriomena furcata 9−Jul 29−Jul 18−Aug 9−Jul 29−Jul 18−Aug 9−Jul 29−Jul 18−Aug 9−Jul 29−Jul 18−Aug 9−Jul 29−Jul 18−Aug Epione vespertaria 9−Jul 29−Jul 18−Aug 7−Sep 9−Jul 29−Jul 28−Aug 9−Jul 29−Jul 28−Aug 9−Jul 29−Jul 28−Aug 9−Jul 29−Jul 28−Aug

Fig. 2 Observed peak flight dates for three Lepidoptera species gothica, the best model was thermal sum modified by photoperiod, versus those predicted under each of four alternative models: Solar for Hydriomena furcata photoperiod, and for Epione vespertaria day (Sday), thermal sum (Tsum), Photoperiod, and thermal sum thermal sum. See Online Resource 6 for classifications of all 112 modified by Photoperiod (Tsum \ Photoperiod). For Orthosia species that were studied 123 Oecologia (2011) 165:237–248 243

Table 2 Frequency (and % of total) of (A) spring, summer or fall ﬂying species and (B) species overwintering as egg, small larva or pupa with different top models for control of ﬂight phenology T TP T,TP P Sday Other

A Spring 1 (11%) 6 (67%) 1 (11%) 0 (0%) 0 (0%) 1 (11%) Summer 25 (29%) 11 (13%) 12 (14%) 14 (16%) 7 (8%) 17 (20%) Fall 0 (0%) 1 (6%) 0 (0%) 13 (76%) 1 (6%) 2 (12%) B Egg 3 (6%) 6 (13%) 4 (9%) 19 (40%) 6 (13%) 9 (19%) Small larva 15 (41%) 3 (8%) 6 (16%) 3 (8%) 2 (5%) 8 (22%) Pupa 6 (32%) 7 (37%) 1 (5%) 2 (11%) 0 (0%) 3 (16%)

T Tsum;TPTsum \ Photoperiod;T,TPTsum or Tsum \ Photoperiod;PPhotoperiod; Sday Solar day comparisons of the goodness of fit for Sday, Tsum, df = 5, p = 0.58; monophagous were excluded having too Tsum \ Photoperiod and Photoperiod models for three few representatives and oligophagous/strongly oligopha- abundant species. gous combined) or among Holarctic, Palearctic and West Palearctic species (v2 = 16.8, df = 10, p = 0.08) or Comparison of species groups between species distributed from South to Central Finland versus species distributed from South to North Finland There was no evidence of phylogenetic constraints in (v2 = 6.5, df = 5, p = 0.26; species restricted to South environmental controls on phenology. Superfamilies Geo- Finland excluded here after having too few representatives). metroidea and Noctuoidea did not differ in the frequencies of species with best model classified as Tsum, Base temperatures and Photoperiodic thresholds Tsum \ Photoperiod, Photoperiod (including P or S), Tsum or Tsum \ Photoperiod, Sday or ‘‘other’’ (including For those 44 species with either a Tsum or Tsum \ Pho- other combinations) (v2 = 7.2, df = 5, p = 0.21 for toperiod as their best model, base temperatures ranged Pearson’s Chi-squared test). Other superfamilies had too between -5 and ?9C (Fig. 3a, b). The distribution of few species to be included in this test. estimated base temperatures was rather uniform for species However, species flying in spring, summer or fall dif- with Tsum and modal (mode at -5C) for species with fered in the frequencies of best models classified as above Tsum \ Photoperiod as their best model. Among the 18 (Table 2a; v2 = 49.4, df = 10, p \ 0.005). The most fre- species having Tsum \ Photoperiod as their best model, quent class of environmental control for spring flyers was the most frequent critical day length was 15 h of daylight Tsum \ Photoperiod (six of nine species). For summer (Fig. 3c). fliers, the most frequent class of environmental control was thermal sum by itself (25 of 86 species), but they also provided cases of photoperiodic control by itself (14 spe- Discussion cies), and many cases of Tsum \ Photoperiod (11 species) and cases where Tsum and Tsum \ Photoperiod could not Strengths and limitations of the modeling procedure be distinguished (12). Among fall fliers, the most common situation was control by photoperiod alone (13 of 17 Our method was an example of using process-based models species). to predict phenological patterns via inverse modeling There were also patterns in phenological controls with (Chuine 2000; Linkosalo et al. 2000, 2006, 2008; respect to overwintering stage (Table 2b; v2 = 34.3, Migliavacca et al. 2008; Morin et al. 2009). Although the df = 10, p \ 0.005). Of those that overwinter as egg the exact nature of the physiological processes controlling environmental controls most frequently involved Photo- the phenology of the studied moth species is unknown, the period (19 of 47 species), of those that overwinter as small models were selected to describe the known physiological larvae Tsum (15 of 37 species) and of those that overwinter processes as closely as possible. as pupa either Tsum \ Photoperiod or Tsum (14 of 19 The strength of our technical approach is that we were species). Other overwintering stages had too few repre- able to screen [100 species for the nature of their pheno- sentatives for evaluation. logical controls. Ours is the first such survey that we know There were no patterns in phenological controls evident of concerning the phenological controls in a diverse insect among oligophagous and polyphagous species (v2 = 3.8, community. It would take a productive research group 123 244 Oecologia (2011) 165:237–248

A Tsum thermal response (commonly experienced in Finland in spring time, in particular, when temperatures typically fluctuate around freezing point) could have contributed to the (surprisingly) high frequency of best models with low base temperatures and late starts in our study. By working with hourly temperatures, we avoided what can be an important bias in the fitting of thermal sum models that 01234567 arises from days when the average temperature is below the −5 0 5 10 Base temperature base temperature but maximum temperatures rise above the base (Ruel and Ayres 1999). Despite their limitations, our B Tsum and Photoperiod results add to the evidence that thermal sum models, sim- plifications that they are, can be applied to diverse taxa to account for considerable variation in phenology.

Adaptive plasticity of moth phenology

It is clear that temperature is a strong proximate determi- 01234567 nant of the annual timing of moth flight in Fennoscandia: Number of Lepidoptera species −5 0 5 10 51% of the studied species showed evidence of thermal Base temperature control. None of the species with Tsum model and only C 28% of the species with Tsum \ Photoperiod model showed population differentiation. When population differences disappear after accounting for temperature (and photoperiod), i.e., when the timing of peak flight varies between sites and years, but the thermal time needed for the peak flight does not, it is logical (and parsimonious) to discount local adaptation and to conclude that these species 012345 have a plastic response to temperatures and can advance 6 8 10 12 14 16 18 their phenology immediately in response to climate Critical day length warming without requirements for evolutionary change. Fig. 3 Frequencies of estimated base temperatures for species with For spring flying species, all which overwinter as pupa, either Tsum (a)orTsum \ Photoperiod (b) identified as the best flight seems to be most frequently controlled by tempera- model and frequencies of critical day lengths for species with tures and photoperiod together: i.e., temperature-dependent Tsum \ Photoperiod identified as the best model (c) development only proceeds after a critical photoperiod is reached. This is generally considered to be adaptive in many years to experimentally study the phenological con- preventing individuals from starting their flight too early in trol systems of so many species. Our results provided years with bouts of warm weather in late winter and early strong evidence of alternative systems of phenological spring. The most frequent critical day-length of moths, controls and variation among species in crucial parameters 15 h of daylight, fits well with the timing of dormancy such as the base temperature for developmental processes. termination in boreal trees (see below). Also, the ‘‘thermic The frequent cases of high goodness-of-fit to theoretically spring’’ (when average daily temperatures rise permanently derived process-based models suggest that we can have to [0C) starts at approximately this time of year (Finnish reasonable confidence in projecting community-wide phe- Meteorological Institute 2009). However, a word of caution nological responses to climate change. This approach does has to be added to the interpretation of the results con- not negate the need for experimental studies. Instead, it cerning spring-flyers. Many of them experience low tem- creates opportunities for focused experimental studies on peratures most part of their spring development and this selected species to further refine our understanding of the part of the thermal response may be strongly non-linear, most common control systems. In this way, the combina- making it potentially difficult to model their phenology tion of inverse modeling and experimental studies is more with linear models. powerful than either could be by itself. In many summer flying species, which generally over- A possible limitation of our approach is the simplifying winter as eggs or small larvae, the flight phenology seems assumption concerning the linearity of the thermal to be generally predictable based on temperatures alone. response (Worner 1992). The non-linear lower end of the This suggests that diapause has ended before temperatures 123 Oecologia (2011) 165:237–248 245 rise above the lower developmental threshold and that sharing the same host (Bryant et al. 2000). In plant com- warm periods early in the season can therefore advance the munities, longer growing seasons (e.g., Myneni et al. 1997; phenology without attenuation from competing physio- Linderholm et al. 2008) but shorter life cycle of individual logical controls. These species should be quite responsive species in response to rising temperatures is most likely to exceptionally warm periods in spring. This could be explained by divergent phenological responses of different risky, because the projected future warming in the northern species in a community (Steltzer and Post 2009). regions of Baltic Sea Basin, including land areas of Fin- Climate change will challenge the ability of individuals land, is estimated to be largest in winter (Graham et al. to time the different life stages appropriately with respect 2008). Some individuals could resume their development to other levels of trophic systems. The more different are too early in the spring, when they could be still vulnerable the underlying physiological controls on phenology among to sudden low temperatures and/or become asynchronized plants, primary consumers, and secondary consumers, the with their host plant. more sensitive will the system be to trophic mismatches. Only one fall flying species showed thermal control of The phenology of budburst, leaf expansion and growth, phenology. The decreased goodness of fit of thermal which are relevant to the nutritional ecology of many models towards the end of summer could be expected due caterpillars (Feeny 1970; van Asch and Visser 2007), also to variation in the quantity or quality of larval food, tend to be temperature-driven in north European tree spe- droughts or different temperature requirements for differ- cies, many of which are important host plants for moths ent physiological processes (e.g., different larval instars) studied here (Ha¨nninen 1995; Linkosalo et al. 2000, 2006, (Danks 1987). On the other hand, all species for which our 2008;Ha¨nninen et al. 2007). Furthermore, the dormancy of results indicated simple photoperiodic control of phenol- many boreal tree species seems to end around March or ogy fly in the late summer (median captures on 20 July or April, which is later in the spring than suggested by after). These species could be capable of adaptive adjust- chilling requirement alone (Linkosalo et al. 2000;Ha¨nninen ments of development that fine tune their growth in et al. 2007) and is likely to be controlled by photoperiod response to photoperiod (Gotthard 2008). Towards late (Myking and Heide 1995; Olsen et al. 1997). The phenology summer or fall, individuals can shorten their juvenile of herbs and other short lived plants, which are also development, for example by metamorphosing at a smaller important hosts for moths, is less studied but they seem to size or speeding up the growth rate. Either could explain be photoperiod-insensitive in spring (Ko¨rner and Basler why the flight in late summer is more synchronized than 2010). what is predicted by thermal control. Yet, this does not The physiological controls of seasonal rhythms in birds, exclude the possibility that the phenology of earlier stages the most important vertebrate predators of mid- to high- of these species, living in spring, can be controlled by latitude Lepidoptera, are known to involve endogenous temperatures. Also, the termination of summer aestivation circannual rhythms and photoperiod (Berthold and Terrill is frequently controlled by photoperiod (Tauber et al. 1991; Hagan et al. 1991), which suggests that they will be 1986). One of our fall-flying species, has a markedly longer generally less responsive to climate change than their prey. pupal duration in southern populations than in northern Accordingly, the limited data that we know of indicate that populations (*12 vs 6 weeks; Tanhuanpaä¨ et al. 1999), warm springs advance the return of migratory birds less which suggests genetic differentiation in a developmental than it advances plant and insect phenology (Marra et al. pause during pupation. 2005; Both et al. 2009). To the extent that migratory birds are less responsive to climate warming than their insect Consequences to communities and trophic systems prey, there will be lighter predation on mid- to high-latitude insects early in the season. The variation among Lepidoptera species in controls of phenology, base temperatures, critical day-lengths, and Alternative models thresholds indicates that there will be differences among species in their responses to climate change because the For 13% of the species, there was plenty of variation in phenology of some species is accelerated by warming more timing of flight among sites and years, but our ensemble of than others. Climate warming could therefore change the possible models was inadequate to capture the true controls temporal structure of the moth community, including tim- on phenology. For these species, more complicated con- ing of flight, timing and duration of larval feeding and trols should be investigated in the future. So far, we have pupation. This could have consequences for competition, only a few experimental studies suggesting what these shared enemies and vacant niches. Differences in thermal could be: nonlinear temperature responses (Peterson and responses and their consequences for phenology and range Nilssen 1996) or complex responses to chilling tempera- limits have been reported previously for four butterflies tures (Ti et al. 2004; Gray et al. 2001). Our models did not 123 246 Oecologia (2011) 165:237–248 consider chilling requirements, because we assumed that thermal time required for development for some species, these are automatically fulfilled in our study region by the they could have missed other cases of differentiation, such as long winters. Other elements that deserve consideration in differences in growth rate, base temperatures or critical day the future analyses include: a critical high or low temper- lengths terminating diapause. There is evidence for variation ature needed as a cue for diapause to end (Tauber et al. in base temperatures within species (Danks 1987; Ayres and 1986; Danks 1987), extreme high or low temperatures Scriber 1994) and many diapause characteristics are known terminating diapause (Danks 1987) or delaying the devel- to vary among populations including the timing of diapause opment (Chown and Terblanche 2007), a certain level of termination (Dambroski and Feder 2007). humidity needed before development can resume after diapause (Danks 1987) and temperature or photoperiodic Conclusions conditions during induction of diapause determining the duration of diapause (Tauber et al. 1986; Danks 1987). Understanding the nature of genetic and physiological Furthermore, part of the unexplained variability in our controls on insect phenology is fundamental to predicting models can be related to varying host plants, since they can the ecological consequences of climate change. Knowledge strongly influence the growth rates of larvae (e.g., Scriber is growing, but even for insects, which must be the best 1981) and therefore timing of adult stage. studied taxa in this respect, much remains a black box. Our study enhances our understanding of the proximate controls Population differentiation on moth phenology but also indicates that more complicated controls probably exist. In the future, experiments One-third of the species recorded from two or more areas linking the herbivore phenology with that of its host are and with Tsum \ Photoperiod model and two-thirds of the needed to better understand the details and differences of species recorded from two or more areas and with Pho- the mechanisms by which herbivores and their hosts track toperiod model ranked as the best model showed popula- the seasons. At the present, we are still in the beginning of tion differentiation, meaning that either the threshold trying to predict and understand when and where trophic thermal sum (Tsum \ Photoperiod) or the critical day systems could become destabilized via climatic effects on length (Photoperiod) for flight was different in different phenology. regions or among populations. Population differentiation in thermal responses may rise from phenotypic plasticity Acknowledgments We thank A. Shapiro and three anonymous (including acclimation) or evolutionary changes (Klok and reviewers for insightful comments on the manuscript. We are grateful to Liisa Tuominen-Roto and Guy So¨derman (Finnish Environment Chown 2003; Chown and Terblanche 2007). Also, the Institute) for their help with the Nocturna database, Matti Rousi and duration of diapause can be modified by the environmental Hanni Sikanen for allowing us to use the hourly temperature data conditions (photoperiod or temperatures) during diapause from Punkaharju and the voluntary Finnish lepidopterists for main- induction (Danks 1987). Furthermore, phenotypic plastic- taining the traps and identifying the moth samples. The study was funded by Emil Aaltonen foundation (A.V.). ity can itself evolve and thus be an expression of local adaptation. Therefore, population differentiation in the controls on phenology, as observed in this study, does not necessarily imply genetic differences among populations. References However, it implies a capacity for adaptive adjustments to climate change: both plastic responses and genetic varia- Altermatt F (2010) Climatic warming increases voltinism in European butterflies and moths. Proc R Soc Lond B 277:1281–1287. doi: tion in photoperiodic or thermal responses (as source 10.1098/rspb.2009.1910 material for selection) help to maintain fitness under Anderson DR (2007) Model based inference in the life sciences. A changing conditions. primer on evidence. Springer, New York Species with a wide distribution have to deal with great Ayres MP, Scriber JM (1994) Local adaptation to regional climates in Papilio canadensis (Lepidoptera: Papilionidae). Ecol Monogr diversity of climatic environments, which can be seen as a 64:465–482 large variation in phenology among geographic populations Berthold P, Terrill SB (1991) Recent advances in studies of bird (Tauber and Tauber 1981). The thermal time required for migration. Annu Rev Ecol Syst 22:357–378. doi:10.1146/annurev. development tends to increase with increasing latitude, es.22.110191.002041 Both C, van Asch M, Bijlsma RG, van den Burg AB, Visser ME although there is much variation in this trend (Honeˇk 1996; (2009) Climate change and unequal phenological changes across Pimentel et al. 2010). For example, in Papilio canadensis, four trophic levels: constraints or adaptations? J Anim Ecol the northern populations of Alaska produce larger eggs, 78:73–83. doi:10.1111/j.1365-2656.2008.01458.x grow faster, and pupate smaller than southern populations to Bradshaw WE, Holzapfel CM (2010) Light, time, and the physiology of biotic response to rapid climate change in animals. Annu compensate for the short growing season (Ayres and Scriber Rev Physiol 72:147–166. doi:10.1146/annurev-physiol-021909- 1994). Although our analyses showed differentiation in 135837 123 Oecologia (2011) 165:237–248 247

Bryant SR, Thomas CD, Bale JS (2000) Thermal ecology of of bud burst with a whole-tree chamber experiment. Tree Physiol gregarious and solitary nettle-feeding nymphalid butterfly larvae. 27:291–300. doi:10.1093/treephys/27.2.291 Oecologia 122:1–10. doi:10.1007/PL00008825 Honeˇk A (1996) Geographical variation in thermal requirements for Cesaraccio C, Spano D, Duce P, Snyder RL (2001) An improved insect development. Eur J Entomol 93:303–312 model for determining degree–day values from daily tempera- Klok CJ, Chown SL (2003) Resistance to temperature extremes in ture data. Int J Biometeorol 45:161–169. doi:10.1007/s00484 sub-Antarctic weevils: interspecific variation, population differ- 0100104 entiation and acclimation. Biol J Linn Soc 78:401–414. doi: Chown SL, Terblanche JS (2007) Physiological diversity in insects: 10.1046/j.1095-8312.2003.00154.x ecological and evolutionary contexts. Adv Insect Physiol Ko¨rner C, Basler D (2010) Phenology under global warming. Science 33:50–152. doi:10.1016/S0065-2806(06)33002-0 327:1461–1462. doi:10.1126/science.1186473 Chuine I (2000) A unified model for budburst of trees. J Theor Biol Kuchlein JH, Ellis WN (1997) Climate-induced changes in the 207:337–347. doi:10.1006/jtbi.2000.2178 microlepidoptera fauna of the Netherlands and the implications Dambroski HR, Feder JL (2007) Host plant and latitude-related for nature conservation. J Insect Conserv 1:73–80 diapause variation in Rhagoletis pomonella: a test for multifac- Kullberg J, Albrecht A, Kaila L, Varis V (2009) Checklist of Finnish eted life-history adaptation on different stages of diapause Lepidoptera—an updated version. http://www.fmnh.helsinki.fi/ development. J Evol Biol 20:2101–2112. doi:10.1111/j.1420- elainmuseo/hyonteiset/perhoset/ [accessed June 2009] 9101.2007.01435.x Leather SR, Walters KFA, Bale JS (1993) The ecology of insect Danilevskii AS (1965) Photoperiodism and seasonal development of overwintering. Cambridge University Press, Cambridge insects. Oliver & Boyd, Edinburgh Linderholm HW, Walther A, Chen D (2008) Twentieth-century trends Danks HV (1987) Insect dormancy: an ecological perspective. in the thermal growing season in the Greater Baltic Area. Clim Biological Survey of Canada. National Museum of Natural Change 87:405–419. doi:10.1007/s10584-007-9327-3 Sciences, Ottawa Linkosalo T, Carter TR, Ha¨kkinen R, Hari P (2000) Predicting spring Davidson J (1944) On the relationship between temperature and rate phenology and frost damage risk of Betula spp. under climatic of development of insects at constant temperatures. J Anim Ecol warming: a comparison of two models. Tree Physiol 13:26–38 20:1175–1182. doi:10.1093/treephys/20.17.1175 Denlinger DL (2002) Regulation of diapause. Annu Rev Entomol Linkosalo T, Ha¨kkinen R, Ha¨nninen H (2006) Models of the spring 47:93–122. doi:10.1146/annurev.ento.47.091201.145137 phenology of boreal and temperate trees: is there something Egerton FN (2006) A history of the ecological sciences. Part 21: missing? Tree Physiol 26:1165–1172. doi:10.1093/treephys/ Reáumur and his history of insects. Bull Ecol Soc Am 26.9.1165 87:212–224. doi:10.1890/0012-9623(2006)87[212:AHOTES]2. Linkosalo T, Lappalainen HK, Hari P (2008) A comparison of 0.CO;2 phenological models of leaf bud burst and flowering of boreal Environment Data Centre (1994) Moth monitoring scheme. A trees using independent observations. Tree Physiol 28:1873– handbook for field work and data reporting. National Board of 1882. doi:10.1093/treephys/28.12.1873 Waters and the Environment, Helsinki Logan JA, Wollkind DJ, Hoyt SC, Tanigoshi LK (1976) Analytic Feeny P (1970) Seasonal changes in oak leaf tannis and nutrients as a model for description of temperature-dependent rate phenomena cause of spring feeding by winter moth caterpillars. Ecology in arthropods. Environ Entomol 5:1133–1140 51:565–581 Marra PP, Francis CM, Mulvihill RS, Moore FR (2005) The influence Finnish Meteorological Institute (2009) Keva¨ttilastot [spring statis- of climate on the timing and rate of spring bird migration. tics]. http://www.fmi.fi/saa/tilastot_50.html#2 [accessed Oecologia 142:307–315. doi:10.1007/s00442-004-1725-x November 2009] Masaki S (1980) Summer diapause. Annu Rev Entomol 25:1–25 Forister ML, Shapiro AM (2003) Climatic trends and advancing Migliavacca M, Cremonese E, Colombo R, Busetto L, Galvagno M, spring flight of butterflies in lowland California. Glob Change Ganis L, Meroni M, Pari E, Rossini M, Siniscalco C, Morra di Biol 9:1130–1135. doi:10.1046/j.1365-2486.2003.00643.x Cella U (2008) European larch phenology in the Alps: can we Gillooly JF, Charnov EL, West GB, Savage VM, Brown JH (2002) grasp the role of ecological factors by combining field obser- Effects of size and temperature on developmental time. Nature vations and inverse modelling? Int J Biometeorol 52:587–605. 417:70–73. doi:10.1038/417070a doi:10.1007/s00484-008-0152-9 Gotthard K (2008) Adaptive growth decisions in butterflies. Biosci- Morin X, Lechowicz MJ, Augspurger C, O’Keefe J, Viner D, Chuine ence 58:222–230. doi:10.1641/B580308 I (2009) Leaf phenology in 22 North American tree species Graham LP, Chen D, Christensen OB, Kjellstro¨m E, Krysanova V, during the 21st century. Glob Change Biol 15:961–975. doi: Meier HEM, Radziejewski M, Raïsa¨nen J, Rockel B, 10.1111/j.1365-2486.2008.01735.x Ruosteenoja K (2008) Projections of future anthropogenic Myking T, Heide OM (1995) Dormancy release and chilling climate change. In: The BACC Author Team (eds) Assessment requirements of buds of latitudinal ecotypes of Betula pendula of climate change for the Baltic Sea Basin. Springer, Berlin, and B. pubescens. Tree Physiol 15:697–704. doi:10.1093/ pp 133–219 treephys/15.11.697 Gray DR, Ravlin FW, Braine JA (2001) Diapause in the gypsy moth: Myneni RB, Keeling CD, Tucker CJ, Asrar G, Nemani RR (1997) a model of inhibition and development. J Insect Physiol Increased plant growth in the northern high latitudes from 1981 47:173–184. doi:10.1016/S0022-1910(00)00103-7 to 1991. Nature 386:698–702. doi:10.1038/386698a0 Hagan JM, Lloyd-Evans TL, Atwood JL (1991) The relationship Nylin S, Gotthard K (1998) Plasticity in life-history traits. Annu Rev between latitude and the timing of spring migration of North Entomol 43:63–83 American landbirds. Ornis Scand 22:129–136 Olsen JE, Junttila O, Moritz T (1997) Long-day induced bud break in Ha¨nninen H (1995) Effects of climatic change on trees from cool and Salix pentandra is associated with transiently elevated levels of temperate regions: an ecophysiological approach to modelling of GA(1) and gradual increase in indole-3-acetic acid. Plant Cell bud burst phenology. Can J Bot 73:183–199 Physiol 38:536–540 Ha¨nninen H, Slaney M, Linder S (2007) Dormancy release of Norway Perrins CM, McCleery RH (1989) Laying dates and clutch size in the spruce under climatic warming: testing ecophysiological models Great tit. Wilson Bull 101:236–253

123 248 Oecologia (2011) 165:237–248

Peterson NA, Nilssen AC (1996) Nonlinear temperature-dependent Steltzer H, Post E (2009) Seasons and life cycles. Science development of autumnal moth pupae, Epirrita autumnata 324:886–887. doi:10.1126/science.1171542 (Lepidoptera: Geometridae). Environ Entomol 25:147–154 Tammaru T, Tanhuanpa¨a¨ M, Ruohoma¨ki K, Vanatoa A (2001) Pimentel C, Ferreira C, Nilsson J-A˚ (2010) Latitudinal gradients and Autumnal moth—why autumnal? Ecol Entomol 26:646–654. the shaping of life-history traits in a gregarious caterpillar. Biol J doi:10.1046/j.1365-2311.2001.00363.x Linn Soc 100:224–236. doi:10.1111/j.1095-8312.2010.01413.x Tanhuanpa¨a¨ M, Ruohoma¨ki K, Kaitaniemi P, Klemola T (1999) Post E, Forchhammer MC (2008) Climate change reduces reproduc- Different impact of pupal predation on populations of Epirrita tive success of an Arctic herbivore through trophic mismatch. autumnata (Lepidoptera; Geometridae) within and outside the Philos Trans R Soc Lond B 363:2369–2375. doi:10.1098/rstb. outbreak range. J Anim Ecol 68:562–570 2007.2207 Tauber CA, Tauber MJ (1981) Insect seasonal cycles: genetics and Post E, Forchhammer MC, Bret-Harte MS, Callaghan TV, Christen- evolution. Annu Rev Ecol Evol S 12:281–308 sen TR, Elberling B, Fox AD, Gilg O, Hik DS, Hoye TT, Ims Tauber MJ, Tauber CA, Masaki S (1986) Seasonal adaptations of RA, Jeppesen E, Klein DR, Madsen J, McGuire AD, Rysgaard S, insects. Oxford University Press, New York Schindler DE, Stirling I, Tamstorf MP, Tyler NJC, van der Wal Ti X, Tuzuki N, Tani N, Morigami E, Isobe M, Kai H (2004) R, Welker J, Wookey PA, Schmidt NM, Aastrup P (2009) Demarcation of diapause development by cold and its relation to Ecological dynamics across the arctic associated with recent time-interval activation of TIME-ATPase in eggs of the climate change. Science 325:1355–1358. doi:10.1126/science. silkworm, Bombyx mori. J Insect Physiol 50:1053–1064. doi: 1173113 10.1016/j.jinsphys.2004.09.007 Pritchard G, Harder LD, Mutch RA (1996) Development of aquatic Trudgill DL, Honeˇk A, Li D, van Straalen NM (2005) Thermal insect eggs in relation to temperature and strategies for dealing time—concepts and utility. Ann Appl Biol 146:1–14. doi: with different thermal environments. Biol J Linn Soc 58:221–244 10.1111/j.1744-7348.2005.04088.x R Development Core Team (2008) R: A language and environment van Asch M, Visser ME (2007) Phenology of forest caterpillars and for statistical computing. R Foundation for statistical computing, their host trees: the importance of synchrony. Annu Rev Entomol Vienna, Austria. http://www.R-project.org 52:37–55. doi:10.1146/annurev.ento.52.110405.091418 Rathcke B, Lacey EP (1985) Phenological patterns of terrestrial van der Have TM (2002) A proximate model for thermal tolerance in plants. Annu Rev Ecol Evol Syst 16:179–214 ectotherms. Oikos 98:141–155 Ruel JJ, Ayres MP (1999) Jensen’s inequality predicts effects of Visser ME, Holleman LJM (2001) Warmer springs disrupt the environmental variation. Trends Ecol Evol 14:361–366 synchrony of oak and winter moth phenology. Proc R Soc Lond Scoble MJ (1995) The Lepidoptera: form, function and diversity. B 268:289–294. doi:10.1098/rspb.2000.1363 Oxford University Press, Oxford Willis CG, Ruhfel B, Primack RB, Miller-Rushing AJ, Davis CC Scriber JM (1981) Sequential diets, metabolic costs, and growth of (2008) Phylogenetic patterns of species loss in Thoreau’s woods Spodoptera eridania (Lepidoptera: Noctuidae) feeding upon dill, are driven by climate change. Proc Natl Acad Sci USA lima bean, and cabbage. Oecologia 51:175–180 150:17029–17033. doi:10.1073/pnas.0806446105 Sharpe PJH, DeMichele DW (1977) Reaction kinetics of poikilotherm Worner S (1992) Performance of phenological models under variable development. J Theor Biol 64:649–670 temperature regimes: consequences of the Kaufmann or rate Stamp NE, Casey TM (eds) (1993) Caterpillars: ecological and summation effect. Environ Entomol 21:689–699 evolutionary constraints on foraging. Chapman & Hall, New York

123