Global Change Biology (2006) 12, 2250–2262, doi: 10.1111/j.1365-2486.2006.01266.x

Lagged effects of North Atlantic Oscillation on spittlebug spumarius (Homoptera) abundance and survival

ANTTI HALKKA*, LIISA HALKKA*w ,OLLIHALKKA*w , KAISA ROUKKAw and J U S S I P O K K I w *Department of Biological and Environmental Sciences, University of Helsinki, PO Box 65, 00014, Helsinki, Finland, wTva¨rminne Zoological Station, University of Helsinki, Helsinki, Finland

Abstract The North Atlantic Oscillation (NAO) is a large-scale pattern of climate variability that has been shown to have important ecological effects on a wide spectrum of taxa. Studies on terrestrial invertebrates are, however, lacking. We studied climate-connected causes of changes in population sizes in island populations of the spittlebug (L.) (Homoptera). Three populations living in meadows on small Baltic Sea islands were investigated during the years 1970–2005 in Tva¨rminne archipelago, southern Finland. A separate analysis was done on the effects of NAO and local climate variables on spittlebug survival in 1969–1978, for which survival data existed for two islands. We studied survival at two stages of the life cycle: growth rate from females to next year’s instars (probably mostly related to overwintering egg survival), and survival from third instar stage to adult. The latter is connected to mortality caused by desiccation of plants and spittle masses. Higher winter NAO values were consistently associated with smaller population sizes on all three islands. Local climate variables entering the most parsimonious autoregres- sive models of population abundance were April and May mean temperature, May precipitation, an index of May humidity, and mean temperature of the coldest month of the previous winter. High winter NAO values had a clear negative effect on late instar survival in 1969–1978. Even May–June humidity and mean temperature of the coldest month were associated with late instar survival. The climate variables studied (including NAO) had no effect on the growth rate from females to next year’s instars. NAO probably affected the populations primarily in late spring. Cold and snowy winters contribute to later snow melt and greater spring humidity in the meadows. We show that winter NAO has a considerable lagged effect on April and May temperature; even this second lagged effect contributes to differences in humidity. The lagged effect of the winter NAO to spring temperatures covers a large area in northern and has been relatively stationary for 100 years at least in the Baltic area. Keywords: Arctic Oscillation, food plants, humidity, , local climate, NAO, North Atlantic Oscillation, Philaenus spumarius, spittlebug, spring temperature

Received 6 July 2005; revised version received 1 May 2006 and accepted 5 June 2006

It is generally described by an index depicting sea level Introduction pressure difference between the Arctic and the subtro- The North Atlantic Oscillation (NAO) is the principal pical Atlantic (Hurrell et al., 2003). When the NAO mode of annual to decadal climate variability in the index is positive, westerly winds bring relatively warm extratropical Northern Hemisphere (Hurrell et al., 2003). winter temperatures into northern Europe. A negative mode of the NAO causes the winters to be relatively Correspondence: Antti Halkka, fax 1 358 9 191 57 694, cold. NAO has been repeatedly connected to global e-mail: antti.halkka@helsinki.fi warming in the Northern Hemisphere (Hurrell, 1995);

r 2006 The Authors 2250 Journal compilation r 2006 Blackwell Publishing Ltd LAGGED EFFECTS OF NAO 2251 a positive trend in NAO was associated with over ratio typical of the phloem-feeding aphids) (Horsfield, half of the winter surface warming in the Eurasia 1978; Raven, 1983). A xylem feeder is very sensitive to towards the end of the last century (Gillett et al., withering of the food plant. P. spumarius nymphs are 2003). Recently, the winter NAO index has been sensitive to desiccation, both directly and indirectly, and decreasing, and overall trends in the past 30 years in this species thus is a useful indicator of variability of the NAO and the closely related Arctic Oscillation (AO) humidity of its meadow habitats. Tapping of xylem by have been weak or nonexistent (Cohen & Barlow, 2005). spittlebugs can have profound negative effects on the A recent analysis of coupled climate models shows, food plants (Carson & Root, 1999). This top-down effect however, that the NAO may intensify with further probably can be exaggerated when the plants are under increases in greenhouse gas concentrations (Kuzmina stress during dry periods. et al., 2005). Our analysis is based on a long-term study (1969– The number of reports on the effects of the NAO on 2005) of archipelago spittlebug populations in the the phytoplankton, planktonic invertebrates and fishes Tva¨rminne section of the Gulf of Finland archipelago is large and increasing. A considerable body of litera- (Halkka et al., 2001). We investigate the effects of NAO ture even exists on the NAO and terrestrial plants and and selected local climate variables on spittlebug abun- vertebrates in northern Europe (Post & Stenseth, 1999; dance for 1970–2005. The same variables are tested in Forchhammer, 2001; Helle et al., 2001; Ottersen et al., models of spittlebug survival, for which we have data 2001; Stenseth et al., 2002, 2003). for a shorter period (1969–1978). A specific question is In their review of terrestrial ecosystem response to to find out how and when climate has strongest effects NAO variability, Mysterud et al. (2003) found that on the populations and how NAO might be connected nothing had been published on the possible relation- to such climate phenomena. ships between terrestrial invertebrates and NAO. This surprising lack of coverage seems to continue till the Materials and methods present day, although Sparks et al. (2005) showed that high NAO-index values were associated with greater The centre of the study area lies at 591500N, 231150E numbers of overseas migration of into (indicated as TV in Fig. 4a). The study islands comprise Britain. a part of the western Gulf of Finland archipelago and Insect populations living in tiny meadows on skerries belong to the Tva¨rminne Zoological Station of the of the Baltic Sea offer good opportunities for investiga- University of Helsinki. Two sampling methods for tions on the role of climate in the life of terrestrial collection of the meadow spittlebugs were used, ‘sweep invertebrate populations. The Gulf of Finland archipe- net’ and ‘minicage’ (described in Halkka & Halkka, lago along the north-eastern side of the Baltic Sea is 1990). Both methods were used in such a way that the characterized by numerous low, small rocky skerries sex and colour phenotype of all the captured adult and a smaller number of larger, wooded islands. There individuals were ascertained on the spot, after which is no tide, but south-westerly gales can occasionally the spittlebugs were released into the herbage. Both raise the water level about 1 m above the long-time methods (but on different islands and time periods) average. Tiny meadows on the skerries are sensitive to were used for estimating the sizes of the populations. changes in the physical environment. The soil layer is The sweep net method was used in the years 1970–2005 quite thin and is easily either desiccated or water- in populations on the islands Gulkobben, Rovholmen logged. and Stora Va¨stra La˚nggrundet (SVL). We used the nets Not many species of plants or are able to until we observed a strong diminution of the catch per cope with the harsh conditions. The meadow spittlebug, sweep. Samples were taken between 09:00 and 16:30 Philaenus spumarius (L.) (Homoptera), is one of the most hours Sampling was performed only when vegetation common and ubiquitous in the meadows of the was dry to minimize differences resulting from the skerries (Halkka & Halkka, 1990). Extensive population vertical distribution of the insects. genetic studies indicate that dispersal between islands The relationship between spittlebug nymph survival, is extremely low (Halkka et al., 2001). This means that NAO and various climate variables was studied by the established populations are not significantly affected minicage method, which even provided an indepen- by migration. dent way to investigate effects of climate on spittlebug The meadow spittlebug has been found globally on abundance in 1969–1978. In the minicage method, the many thousands of food plant species and is certainly spittle mass with a P. spumarius nymph was enclosed one of the world’s most extreme polyphages. P. spumar- in a small round styrox box on the food plant. On the ius is a xylem feeder, and can ingest xylem sap 150–200 insides of the box halves, superlon rings were pressed times the body weight over 24 h (about 10 times the against the plant stem. The box was ventilated by nylon r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2250–2262 2252 A. HALKKA et al. windows. The spittlebug nymph emerged into adult the so-called extended winter (December–March) ver- in the box in about 2 weeks. The minicage method was sion of the index. used in the years 1969–1978 on the islands Allgrundet We tested for time lags between independent vari- and Porskobben. ables and population size with cross correlation func- Isolating nymphs in the very conspicuous spittle tion (CCF). Significant effects were found only for lag 0 masses was much more effective than collecting with except for population sizes which were autoregressive the sweep nets. Meadows were visited at least twice to at lag 1. Cross-correlations indicate spatial synchrony ascertain that practically all nymphs could be isolated among local populations. It has been known for a long in cages. Most of the nymphs were at the third instar time that populations may be synchronized by but smaller numbers were at the second or fourth instar climate (Moran, 1956; Ranta et al., 1999; Post & For- stage when placed in the minicages. We examined chammer, 2002). All the significant cross-correlations mortality from the third instar stage to adult stage shown would remain at least at significance level and the relationship between the previous years female Po0.05 after Bonferroni correction. population size and the number of third instars. The We then developed a set on candidate linear first latter sums up fecundity from the previous year, egg order autoregressive models of climate effects on spit- mortality and early instar mortality. This growth ratio tlebug abundance. These models were of the general from females to instars of the next year is probably form: mostly linked to factors of winter climate. The univol- tine spittlebug overwinters at the egg stage. In Finland, Nt ¼ Nðt1Þ þ b0 þ b1NðtÞ þ b2x2ðtÞþþbmxmðtÞþe: nymphs are to be found from late May to July, and adults are seen from July to October (Halkka et al., We selected the most parsimonious models on the 1967). The amplitude of population fluctuations was basis of the Akaike information criterion (AIC), at the same level in the sweep net and minicage corrected for small sample size (AICC, Burnham & methods, indicating that isolation in minicages did not Anderson, 2002). The population sizes were log- cause much excess mortality. transformed and the mortality values arcsin trans- One of the field stations of the Finnish Meteorological formed before analysis. We also used differenced ver- Institute (FMI) is situated at the Tva¨rminne Zoological sions of the climate variables in the models; differencing Station. Data on snow cover, temperature and precipita- is a powerful statistical method for trend elimination. tion in Tva¨rminne were obtained from the FMI. Year was included in the analysis of Stora Va¨stra Bale et al. (2002) regard temperature as the dominant La˚nggrundet, as this population displayed a significant abiotic factor affecting populations of herbivorous in- trend (Post & Stenseth, 1998). As autocorrelation of sects. We used both winter and spring and early sum- time-series can lead to biased estimates of significance, mer temperatures in models predicting spittlebug we adjusted degrees of freedom according to Bartlett abundance and survival. We were specifically inter- (1946), see Post & Stenseth (1998). Many of our expla- ested in the role of humidity as field observations had natory variables were correlated among each other shown desiccation of the habitat meadows to be the (multicollinear). Ecological data may be typically multi- major mortality factor at late nymph stage. A meadow collinear (Graham, 2003), which is a problem as strong humidity index (MHI), gave a rough estimate of soil collinearity can result in unstable models. We used humidity. This simple index consists of precipitation variance inflation factor (VIF) to check for collinearity. sum (in millimeter) from which the temperature sum of As severe collinearity (VIF410) was found between the same period is subtracted. Low temperatures and different versions of the NAO index, we rejected mod- rainfall increase the value of the MHI. The local climate els with more than a single index version from further variables used contain two winter proxies (mean analysis (Graham, 2003). Collinearity was weak temperature of the coldest month and length of the (VIFo2) in all of the most parsimonious models. snowy season), monthly and bimonthly spring–early Statistical analysis was performed with R software summer (April–June) temperatures and precipitation, (version 2.1.1., Ihaka & Gentleman, 1996). and monthly and bimonthly versions of the simple humidity index. Results The NAO index values were obtained from the Cli- matic Research Unit of the University of East Anglia Significant spatial synchrony was found in the study (http://www.cru.uea.ac.uk). This version of the NAO populations in 1970–2005 and to a lesser extent in index (Jones et al., 1997) is based on pressure differences 1970–1978. Spittlebug abundance in all three 35-year between Gibraltar and Iceland. We used a bimonthly study populations was significantly correlated with the (January–February), trimonthly (January–March), and two other populations at lag 0 (Rovholmen–SVL,

r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2250–2262 LAGGED EFFECTS OF NAO 2253 r 5 0.70, t 5 5.7, Po0.001; Rovholmen-Gulkobben, two and seven (median 4.8 for Allgrundet, 3.0 for r 5 0.68, t 5 5.3, Po0.001; Gulkobben–SVL, r 5 0.48, Porskobben) but was once in Allgrundet and once in t 5 3.1, Po0.01). Allgrundet was significantly correlated Porskobben so low (less than two assuming equal with SVL (r 5 0.88, t 5 4.9, Po0.01) and Rovholmen proportion of sexes) that this alone necessitated a drop (r 5 0.86, t 5 4.4, Po0.01), but not with Gulkobben and in population size. Low female density tended to result Porskobben in 1970–1978. For Porskobben, the correla- in proportionally more instars next year, but this asso- tions with SVL, Rovholmen and Gulkobben were not ciation was not statistically significant (Table 4). significant. Gulkobben was not correlated with Rovhol- As nymph mortality was clearly linked to climate men or SVL in 1970–1978, but the correlation between variables, we wanted to check if this was reflected in Rovholmen and SVL was significant also in 1970–1978 abundances, and tested the Porskobben and Allgrundet (r 5 0.90, t 5 5.5, Po0.01). 10-year data with the autoregressive models used for Many of the climate parameters were associated with the 35-year study islands. For Allgrundet, the best each other and the NAO index in 1970–2005; the corre- model contained the population of the previous year lations are not produced by trends as they are present and January–February NAO. For Porskobben, NAO also in differenced series (Table 1). High NAO values was not among the supported models. The highest and high winter and April–May temperatures had a ranking climate variable for Porskobben was minimum negative correlation with population sizes. Large num- temperature of the coldest month, a variable correlated bers of snowy days and high humidity (MHI) values with NAO (1969–1978, r 5 0.66, P 5 0.05, 1971–2005, had a positive association with population size. Asso- r 5 0.69, Po0.001, both for January–February NAO). ciations between selected single variables are shown in As spring temperatures were clearly important for Fig 1. The covariability of normalized climate variables survival and abundance of spittlebugs, we studied the and the pooled 35-year population size of the three spatial extent and time stationarity of the lagged effect study islands is noticeable especially if shown as 3-year of NAO on late spring temperature. The Tva¨rminne moving averages (Fig. 2a and b). temperature records do not extend further back than the The AIC-based rankings of top models of spittlebug 1960s. We studied time-stationarity with temperature abundance are shown in Table 2. The only variable that data of Svenska Ho¨garna (59.271N, 19.31E) in the entered the best model of all three islands (apart from Stockholm archipelago in 1890–2001. This data was the autoregressive population size of the previous year) obtained from the quality-controlled Nordklim data was the NAO index, with the January–February NAO set (Tuomenvirta et al., 2001). A running 21-year corre- present twice and January–March NAO once (Gulkob- lation between January–March NAO and January– ben). According to Burnham & Anderson (2002) models March and April–May temperatures (Fig. 3) shows that that have AIC values that differ from the lowest value the lagged effect of winter NAO is relatively stationary by less than two have ‘substantial support’. Other in Svenska Ho¨garna and stays mostly at about the same parameters entering such top models were minimum level as the direct effect on winter temperatures. A high mean temperature of the coldest month, mean tempera- and persistent correlation (not shown) between winter ture differences of April–May, and the humidity index NAO and April–May mean temperature was obtained of May. Also the components of this index, May tem- also for Falsterbo (south-western Baltic 581N161E) and perature and precipitation, had support. Holmo¨ (northern Baltic, 651N301E, locations shown in Models of nymph survival from third instar to adult Fig. 4a). (late spring–early summer survival) in 1969–1978 are Gridded temperature data of NCEP/NCAR (Kalnay shown in Table 3. January–February NAO was involved et al., 1996) and imaging tools from the NOAA-CIRES in the top models of both Porskobben and Allgrundet. Climate Diagnistics Center (www.cdc.noaa.gov/) were For Porskobben the single model with substantial sup- used to study the spatial dimension of the lagged effect port included NAO and length of the snowy season. For in 1970–2003. Considerable correlation between Janu- Allgrundet NAO and May–June humidity entered the ary–February NAO and April–May temperature is pre- top model, but also models with NAO only as a sent in an area extending from the North Sea to the predictor variable, and May–June humidity together Baltic and Barents Seas (Fig. 4a). with mean temperature of the coldest month, had support. Discussion Survival was not related to initial nymph density, and we could not find any effect of NAO or any of the Atmospheric effects climate variables on the combined egg and early instar survival plus fecundity (Table 4). The number of instars The sensitivity of the spittlebug to the variability of the per female of the previous year stayed mostly between climate has been investigated earlier in Europe and the r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2250–2262 2254 Table 1 Correlation matrix (100) of main variables used in the study (1971–2005). (A) Nondetrended time-series (N 5 35). (B) Difference-detrended (N 5 34)

Stora HALKKA A. V-La˚ng. Rovholmen (SVL) Gulkobben April May June AM MJ PMay Mhum MJhum MinT Snow JFNAO

(A) SVL 70 Gulkobben 68 48 April temperature 35 21 12 al et

May temperature 47 30 24 38 . June temperature 09 08 04 11 10 April–May temperature 49 30 22 84 82 13

ora compilation Journal May–June temperature 34 24 11 31 66 82 57 May precipitation 29 38 22 40 38 18 47 35 May humidity 48 38 27 45 95 14 83 66 65 May–June humidity 37 35 13 31 60 79 54 94 48 65 Minimum monthly winter temperature 46 51 06 44 50 31 57 53 40 55 51 Snowy days 40 33 13 56 55 15 67 44 39 58 42 69 January–February NAO 53 51 27 71 53 24 75 49 38 57 48 69 58 r January–February–March NAO 50 40 33 60 52 23 67 48 34 54 48 52 47 91

06BakelPbihn Ltd, Publishing Blackwell 2006 (B) SVL 76 Gulkobben 64 56 April temperature 41 40 26 May temperature 46 45 18 27 Jun temperature 24 04 07 27 25 April–May temperature 55 53 27 76 84 32 May–June temperature 41 25 14 35 69 88 66 May precipitation 36 43 29 35 55 27 57 48 May humidity 48 48 23 33 97 28 85 70 73 May–June humidity 41 25 15 33 62 88 61 97 51 65 lblCag Biology Change Global Minimum monthly winter temperature 39 44 04 32 62 40 60 61 27 59 51 Snowy days 37 23 12 39 52 17 57 39 18 48 30 63 January–February NAO 49 53 32 61 47 41 67 55 36 49 49 65 40 January–February–March NAO 49 39 33 54 42 41 59 52 34 44 50 49 30 93 r Note that first-order differencing of log population sizes amounts to population growth rate. 06TeAuthors The 2006 NAO, North Atlantic Oscillation. , 12, 2250–2262 LAGGED EFFECTS OF NAO 2255

7.5

7.5 7.0 t

e 6.5 d n u 6.5 r 6.0 g l l A

n

l 5.5 ln Rovholmen 5.5 5.0 (a) (b) 4.5 4.5

–2 –1 0 1234 –1.0 0.0 0.5 1.0 1.5 2.0 January–February NAO January–February NAO

8 8.0 (d)

7.5 7 7.0

6.5 6 6.0 (degrees C) ln Rovholmen 5.5 5

5.0 (c) AprilMay mean temperature 4 4.5 4 5 6 78 −2 −1 01234 April − May mean temperature (degrees C) January− February NAO

(f) 0.8 0.8

0.6 0.6

0.4 0.4

0.2 (e) 0.2 Nymphs surviving to adult stage Nymphs surviving to adult stage

–2 –1 0 1 2 –700 –650 –600 –550 January–February NAO May–June humidity

Fig. 1 Relationships between (a) Rovholmen population size and January–February NAO, (b) Allgrundet population size and January– February NAO. (c) Rovholmen population size and April–May mean temperature in Tva¨rminne, (d) April–May temperature and January–February NAO, (e) Nymph survival to adult stage and January–February NAO, and (f) Nymph survival and May–June humidity. In (e and f), Allgrundet has open and Porskobben filled symbols.

United States. Weaver & King (1954) stressed the sig- that of the most competitive model. The supported nificance of humidity in their extensive and excellent models include spring and early summer variables monograph on P. spumarius. Among 63 insect species for (April–May temperature and May humidity) and mean which Stiling (1988) listed the key factors governing temperature of the coldest month of the previous year. mortality, P. spumarius was one of the six primarily NAO and largely the same climate variables were affected by variable weather (see also Whittaker, 1973). involved in both the 1970–2005 (sweepnet) and 1969– Our results show clearly that NAO had a significant 1978 (minicage) datasets. All of these climate variables effect on meadow spittlebug abundance in Tva¨rminne. were correlated with NAO (Table 1). As NAO stays in Several other variables are also supported as they are the models in their presence, it is clear that the effect of included in models with only slightly larger AIC than the NAO operates only partially through them. r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2250–2262 2256 A. HALKKA et al.

(a) 10

8

6

4

2 INV-NAO SNOW 0 HUMIDITY INV-TEMP PHIL-POP –2

–4

–6

–8

–10 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005

(b) 6

4

2

INV-NAO SNOW 0 HUMIDITY INV-TEMP PHIL-POP

–2

–4

–6 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Fig. 2 Standardized time series of pooled population size of the three 35-year study populations (PHIL-POP), April–May mean temperature (INV-TEMP), May humidity (HUMIDITY), number of snowy days (SNOW) and January–February NAO (INV-NAO). (a) On a yearly basis, (b) smoothed as 3-year moving averages.

It became clear in the 1990s that large-scale indexes climate (where snow cover and minimum temperature like NAO are often better predictors of ecological phe- of the coldest month were used as proxies) affects nomena than local climate variables. The reason for this spring hydrology via snow climate; our study area is may be that NAO combines features of several weather in the section of northern Europe where snow melt components (Stenseth & Mysterud, 2005). Our study dates have a significant correlation with the AO (Schae- indicates, that NAO influences spittlebug abundance fer et al., 2004). The lagged effect of the NAO on April primarily via late spring–early summer nymph mortal- and May temperatures and May precipitation adds to ity, which had a clear association with NAO in 1969– this snow-mediated effect. If the relative importance of 1978, with NAO alone explaining about two-thirds local climate variables differs from year to year, the of the variance in survival (Table 4). This mortality is effect of NAO emerges as the most consistent one. caused by desiccation, which the NAO may influence in As migration is low in the Tva¨rminne spittlebug many ways. The direct influence of NAO on winter populations, NAO associated climate factors were prob-

r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2250–2262 LAGGED EFFECTS OF NAO 2257

Table 2 Performance of models of spittlebug abundance (1971–2005) on the three islands studied

2 Stora Va¨stra La˚nggrundet R AICC DAICC b0 1 b1(Nt1) 1 b2 (year) 1 b3 (JFNAO) 1 b4 (May humidity*) 0.67 61.38 0 b0 1 b1(Nt1) 1 b2 (year) 1 b3 (JFNAO) 1 b4 (mintemp.) 0.67 62.05 0.67 b0 1 b1(Nt1) 1 b2 (year) 1 b3 (JFNAO) 0.63 62.48 1.10 b0 1 b1(Nt1) 1 b2 (year) 1 b3 (JFNAO) 1 b4 (May precipitation) 0.66 62.57 1.19 b0 1 b1(Nt1) 1 b2 (year) 1 b3 (JFNAO) 1 b4 (May temperature*) 0.66 62.73 1.35 b0 1 b1(Nt1) 1 b2 (year) 1 b3 (JFNAO) 1 b4 (April–May temperature*) 0.66 63.01 1.63 Rovholmen b0 1 b1(Nt1) 1 b2 (JFNAO) 1 b3 (April–May temperature*) 0.47 72.9 0 b0 1 b1(Nt1) 1 b2 (JFMNAO) 1 b3 (April–May temperature*) 0.47 73.25 0.35 b0 1 b1(Nt1) 1 b2 (JFMNAO) 1 b3 (May temperature*) 0.46 73.32 0.42 b0 1 b1(Nt1) 1 b2 (JFMNAO) 1 b3 (May humidity*) 0.46 73.50 0.60 b0 1 b1(Nt1) 1 b2 (JFNAO) 1 b3 (May temperature*) 0.46 73.57 0.67 b0 1 b1(Nt1) 1 b2 (JFNAO) 1 b3 (May humidity*) 0.46 73.59 0.69 b0 1 b1(Nt1) 1 b2 (April–May temperature*) 0.41 73.93 1.03 Gulkobben b0 1 b1(Nt1) 1 b2 (JFMNAO) 0.28 74.96 0 b0 1 b1(Nt1) 1 b2 (JFNAO) 0.26 75.85 0.89 b0 1 b1(Nt1) 1 b2 (DJFMNAO) 0.26 76.15 1.19 b0 1 b1(Nt1) 1 b2 (JFNAO) 1 b3 (mintemp.) 0.30 76.69 1.73 b0 1 b1(Nt1) 1 b2 (JFMNAO) 1 b3 (mintemp.) 0.30 76.79 1.83 b0 1 b1(Nt1) 1 b2 (JFMNAO) 1 b3 (May humidity) 0.29 77.28 2.32

NAO is the North Atlantic Oscillation, humidity is a simple index describing hydrology (precipitation-temperature), mintemp. is mean temperature of the coldest month of the previous winter. *Use of differenced series. Models are ranked according to Akaike information criterion (AIC); the best model has a DAIC value of 0. R2 is the proportion of variation explained by the model. Significant variables (Po0.05) in the models are highlighted in bold. Significance levels of independent variables have been adjusted for autocorrelation. ably acting as synchronizing agents. The role of climate lations to NAO apart from the study of Sparks et al. in synchronising population fluctuations has been dis- (2005) on migratory Lepidoptera. The variability of cussed since the 1950s (Moran, 1956, Ranta et al., 1999, El Nin˜o Southern Oscillation is known to exert its Engen et al., 2005), and the NAO has been shown to NAO-like effects in the ecosystems of the North and contribute to synchrony in several vertebrate species South American continents (Holmgren et al., 2001; Lima (see references in Engen et al., 2005). & Jaksic, 2004). In the western–midwestern part of the The models of spittlebug abundance explained from United States, Vandenbosch (2003) reported fluctua- about a third (Gulkobben) to two thirds (Stora Va¨stra tions of Vanessa cardui (Lepidoptera) abundance syn- La˚nggrundet) of the variance of population size. The chronized with oscillations of El Nin˜o and also showing unexplained part of the variation is partly attributable coupling to variability in the Pacific Decadal Oscillation. to sampling error, but there was an important part of The lagged effect of NAO/AO on April and May the life-cycle (combined fecundity, egg mortality and climate has been noticed earlier by Buermann et al. early nymph mortality) for which we could not show (2003), who showed a clear correlation of 1982–1998 any climate covariates. This may be a good explanation January–March AO and March–May temperatures in for NAO not being included in the supported model of northern Europe. Gormsen et al. (2005) linked the April– spittlebug abundance in Porskobben. Our results are of May flowering time of two tree species in Denmark course not sufficient to show that NAO does not affect to December–February NAO. The effect was according winter mortality. We have survival data only for a to them operating both through direct NAO effect on limited period of time, and NAO might have contrast- winter temperatures and a lagged effect of winter NAO ing effects on egg and early nymph survival. The results on March–May temperatures. do indicate, however, that the principal NAO effect is Recently, Menzel et al. (2005) have shown that operating during spring. high NAO winter and early spring influences plant We are not aware of other published studies linking phenology in Europe along the SW–NE axis in the insect species abundances or synchrony of insect popu- beginning of the growing season. This is precisely r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2250–2262 2258 A. HALKKA et al. the period which is vital for success of the P. spumarius plant or animal phenology. Long distance bird nymphs. As the lagged spring effect of winter migration in northern Europe is a good example. NAO covers spatially much of northern Europe and Ornithologists have been puzzled when arrival dates appears to be relatively stationary, it could be contribut- of migrants from Africa have shown relatively strong ing to many of the cases where NAO has been correlations with the winter NAO in as late as May in shown to have an effect on, for example, late spring the north European bird stations Helgoland (North Sea, Hu¨ ppop & Hu¨ ppop, 2003), Ottenby (Baltic, Stervander et al., 2005), and Turku/Jurmo and Hanko Table 3 Performance of models of spittlebug late instar (Baltic, Va¨ha¨talo et al., 2004). Attempts to explain these nymph survival (1969–1978) on Allgrundet and Porskobben correlations with variability in the African climate 2 R DAICC associated with the NAO have not been successful

Allgrundet

b0 1 b1 (May–June humidity) 1 b2 0.81 0 (January–February NAO) 1.0

b0 1 b1 (May–June humidity) 1 b2 0.79 0.95 (mintemp.) 0.8 b0 1 b1 (January–February NAO) 0.63 1.38

b0 1 b1 (May–June humidity) 0.56 2.27 Porskobben 0.6

b0 1 b1 (January–February NAO) 1 b2 0.91 0 (snow) 0.4 b0 1 b1 (January–February NAO) 0.73 5.35 b0 1 b1 (January–February NAO) 1 b2 0.84 6.39 (mintemp.) 0.2

NAO is the North Atlantic Oscillation, humidity is a simple index describing hydrology (precipitation-temperature), min- 0.0 temp. is mean temperature of the coldest month. Models are 1900 1920 1940 1960 1980 ranked according to Akaike information criterion (AIC); the 2 best model has a delta AIC value of 0. R is the proportion of Fig. 3 Running correlation coefficient (21-year window) be- variation explained by the model. All variables in the models tween January–March NAO and April–May (striped) and Jan- were significant (Po0.05) also after adjusting for autocorrela- uary–March temperatures in Svenska Ho¨garna, Stockholm tion. archipelago (SH in Fig. 4a).

Table 4 Correlations of selected climate variables with nymph and overwinter survival (modeled as female to nymph growth rate) in Allgrundet and Porskobben (1969–1978)

Nymph survival Winter survival

Allgrundet late Porskobben late Allgrundet Porskobben

R2 PR2 PR2 PR2 P

Minimum monthly temperature 0.47 o0.05 0.59 o0.01 0.04 0.59 0.03 0.65 Number of snowy days 0.29 0.11 0.37 0.06 0.00 0.85 0.03 0.65 April mean temperature 0.41 o0.05 0.41 0.05 0.00 0.91 0.13 0.34 May mean temperature 0.18 0.22 0.26 0.16 0.06 0.53 0.10 0.39 April–May mean temperature 0.41 o0.05 0.50 o0.05 0.02 0.74 0.00 0.99 May precipitation 0.26 0.13 0.49 o0.05 0.00 0.93 0.18 0.25 May humidity 0.31 0.09 0.49 o0.05 0.06 0.54 0.01 0.76 May–June humidity 0.56 o0.01 0.30 0.10 0.33 0.10 0.00 0.89 January–February NAO 0.60 o0.01 0.73 o0.01 0.00 0.89 0.08 0.47 Initial nymph density 0.04 0.57 0.07 0.48 –––– Initial female density ––––0.38 0.08 0.33 0.11

Correlations with initial nymph and female density are shown. Significant variables (Po0.05, adjusted for autocorrelation) are highlighted in bold. NAO, North Atlantic Oscillation.

r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2250–2262 LAGGED EFFECTS OF NAO 2259

(a)

HO

JU TV

SH

OB

HE FA

(b)

Fig. 4 (a) Lagged correlation between January–February NAO index and gridded April–May temperature in 1971–2000 (Kalnay et al., 1996). TV, Tva¨rminne; SH, Svenska Ho¨garna; HO, Holmo¨; FA, Falsterbo; JU, Jurmo bird station; OB, Ottenby bird station; HE, Helgoland bird station. (b) direct correlation between January–March NAO and gridded January–March surface temperature.

(see Stervander et al., 2005 for a discussion). As the unexpected association of winter NAO with late spring lagged correlation between winter NAO and spring migration. temperatures is strong at all of these North European We suggest that future studies of NAO/AO effects bird stations (Fig 4a), it may be sufficient to explain the should in northern Europe always consider the possible r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2250–2262 2260 A. HALKKA et al. role of lagged effects of winter NAO on spring tempera- The responsiveness of this species to climatic variables ture, and in some cases even lagged summer effects. may well be representative for insects in a broad A lagged positive effect of NAO on summer temperatures manner as temperature and precipitation are dominant has been shown for Britain in 1970–2001 (Qian & Saun- abiotic factors affecting herbivorous insects (Bale et al., ders, 2003). This effect is, however, not stationary in time. 2002). Specifically, plant-mediated indirect effects on The effects of biotic factors other than plant condition climate–insect interactions may be an important me- on spittlebugs may be quite weak in the environment chanism with which climate and global climate change studied, and so do not mask the climate signal shown affects insect population dynamics. Many of the vari- in this study. Predation appears to be low in many ables affecting spittlebug abundance and survival here European and North American populations of P. spu- discussed with NAO are affected by climate change. marius (Whittaker, 1970; Halkka & Kohila, 1976; Harper Climate models predict large changes in temperature, & Whittaker, 1976; Karban, 1989; Halkka & Halkka, precipitation and winter climate in Fennoscandia (e.g. 1990). The albovenosa, Melanchra pisi and Carter et al., 2004). One must, however, bear in mind Proxenus lepigone are possible competitors but they that the Baltic Sea archipelago environment studied occur in very small numbers only, and Simyra uses may, in some aspects, be more sensitive to climatic , a food plant practically never used fluctuations than many mainland environments. by P. spumarius (Gillerfors, 1966; Silfverberg, 1968). Voles may occasionally affect meadows (Halkka et al., 1975), but they did not affect our study populations. Acknowledgements Weaver & King (1954), in Wooster, Ohio, estimated that Warm thanks are due to the personnel of the Tva¨rminne Zool- 10% or less of P. spumarius eggs were parasitized. ogical Station for placing at our disposal all the facilities needed. Harper & Whittaker (1976) reported from an English Finnish Meteorological Institute provided climate data. Professor spittlebug population endoparasitization of 46% (male) Esa Ranta made very useful comments. We thank Riitta Hovi- and 31% (female) adult P. spumarius specimens by the nen, MA, Anne Morikka, MA, Marja-Riitta Eskola, MA, Merja Salmitie, MA, Anita Mu¨ ller (Dresden, Germany), and Vesa, Erkki Pipunculid Falle´n (Diptera). The pipun- and Sara Halkka for technical assistance, and four anonymous culid appear not to attack nymphs (Whit- referees for constructive criticism of an earlier draft of this paper. taker 1969, 1970). In Finland, rates of parasitization of The study was aided by grants from the University of Helsinki eggs have not been recorded. Verrallia aucta is not and the Oscar O¨ flund and Maj and Tor Nessling Foundations. commonly found in the region, and we have found no parasitoids (several hundreds of specimens investi- gated in 2004 and small numbers P. spumarius adults References opened earlier for chromosome studies). It seems that Bale JS, Masters GJ, Hodginson ID et al. (2002) Herbivory in our biotopes in many ways liken those in the Pennine global climate research: direct effects of rising temperature on Hills Moor House nature reserve in England, where insect herbivores. Global Change Biology, 8, 1–16. V. aucta is absent and population sizes of spittlebugs Bartlett MS (1946) On the theoretical specification of sampling are primarily affected by climate (Whittaker, 2001). properties of autocorraleted time series. Journal of the Royal Statistical Society Supplement, 8, 27–41. We did not try to sort out the relative roles of Buermann W, Anderson B, Tucker CJ et al. (2003) Interannual density-dependent and density-independent factors covariability in Northern Hemisphere air temperatures and (see Hunter & Price, 1998; Ranta et al., 2000 about the greenness associated with El Nin˜o-Southern Oscillation and difficulties involved). We assume that possible density the Arctic Oscillation. Journal of Geophysical Research, 108, 4396, dependence may primarily affect fecundity plus early doi: 10.1029/2002JD002630. stages of the life-cycle. Females routinely produced Burnham KP, Anderson DR (2002) Model Selection and Multimodel close to 10–30 third instars (maximum 94) in our cross- Interference: A Practical Information-Theoretic Approach, 2nd edn. ing experiments (unpublished data). In this study, Springer-Verlag, New York. female to next year third instar ratio exceeded 10 in Carson WP, Root RB (1999) Top-down effects of insect herbivores Porskobben (1973–1974) and Allgrundet (1971–1972). It during early succession: influence on biomass and plant is clear that such high productivity has its limits in the dominance. Oecologia, 121, 260–272. Carter TR, Fronzek S, Barlund I (2004) FINSKEN: a framework long term with carrying capacity possibly operating at ¨ for developing consistent global change scenarios for Finland an early instar stage. But it is evident that the skerry in the 21st century. Boreal Environment Research, 9, 91–107. populations may face climate-driven crashes of popula- Cohen J, Barlow M (2005) The NAO, the AO, and global warm- tion size before density dependent forces begin to have ing: how closely related. Journal of Climate, 18, 4498–4513. considerable effect (White, 2004). Engen S, Lande R, Saether B-E et al. (2005) Estimating the pattern P. spumarius is one of the most abundant insects in the of synchrony in fluctuationg populations. Journal of Animal boreal and temperate biomes (Halkka & Halkka, 1990). Ecology, 74, 601–611.

r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2250–2262 LAGGED EFFECTS OF NAO 2261

Forchhammer MC (2001) Terrestrial ecological responses to Climatic Significance and Environmental Impact (eds Hurrell JW, climate change in the Northern Hemisphere. In: Climate change Kushnir Y, Ottersen G, Visbeck M). Geophysical Monograph, 134, research – Danish contributions (eds Jorgensen AMK, Fenger J, 1–35. Halsnaes K), pp. 219–236. GAD, Copenhagen. Ihaka R, Gentleman R (1996) R: a language for data analysis and Gillerfors G (1966) Insektfaunans zonering i O¨ stergo¨tlands ska¨r- graphics. Journal of Computational and Graphical Statistics, 5, ga˚rd. Coleoptera and Heteroptera. Opuscula Ento- 299–314. mologica, Supplement, 31, 1–124. Jones PD, Jo´nsson T, Wheeler D (1997) Extension to the North Gillett NP, Graff HF, Osborn TJ (2003) Climate change and the Atlantic Oscillation using early instrumental pressure obser- NAO. In: The North Atlantic Oscillation. Climatic Significance and vations from Gibraltar and South–West Iceland. International Environmental Impact (eds Hurrell JW, Kushnir Y, Ottersen G, Journal Climatology, 17, 1433–1450. Visbeck M). Geophysical Monograph, 134, 193–209. Kalnay E, Kanamitsu M, Kistler R et al. (1996) The NCEP/NCAR Gormsen AK, Hense A, Toldam-Andersen TB et al. (2005) Large- 40-year reanalysis project. Bulletin of the American Meteorologi- scale climate variability and its effects on mean temperature cal Society, 77, 437–471. and flowering time of Prunus and Betula in Denmark. Theore- Karban R (1989) Community organization of Erigeron glaucus tical and Applied Climatology, 82, 41–50. folivores: effects of competition, predation and host plant. Graham MH (2003) Confronting multicollinearity in ecological Ecology, 70, 1028–1039. multiple regression. Ecology, 84, 2809–2815. Kuzmina SI, Bengtsson L, Johannessen OM et al. (2005) The Halkka O, Halkka L (1990) Population genetics of the poly- North Atlantic Oscillation and greenhouse gas forcing. Geo- morphic meadow spittlebug, Philaenus spumarius (L.). Evolu- physical Research Letters, 32, 1–4. tionary Biology, 24, 149–191. Lima M, Jaksic FM (2004) The impacts of ENSO on terrestrial Halkka O, Halkka L, Roukka K (2001) Selection often overrides ecosystems: a comparison with NAO. In: Marine Ecosystems the effects of random processes in island populations of and Climate Variation (eds Stenseth NC, Ottersen G), pp. 169– Philaenus spumarius (Homoptera). Biological Journal of the Lin- 175; 224–227. Oxford University Press, Oxford, UK. nean Society, 74, 571–580. Menzel A, Sparks TH, Estrella N et al. (2005) ‘SSW to NNE’ – Halkka O, Kohila T (1976) Persistence of visual polymorphism, North Atlantic Oscillation affects the progress of seasons despite a low rate of predation, in Philaenus spumarius (L.) across Europe. Global Change Biology, 11, 909–918. (Homoptera, ). Annales Zoologici Fennic, 13, Moran P (1956) The statistical analysis of the Canadian lynx 185–188. cycle II. Synchronization and meteorology. Australian Journal Halkka O, Raatikainen M, Halkka L et al. (1975) The genetic Zoology, 1, 291–298. composition of Philaenus spumarius populations in island Mysterud A, Stenseth NC, Yoccoz NG et al. (2003) The response habitats variably affected by voles. Evolution, 29, 700–706. of terrestrial ecosystems to climate variability associated with Halkka O, Raatikainen M, Vasarainen A et al. (1967) Ecology and the North Atlantic Oscillation. In: The North Atlantic Oscilla- ecological genetics of Philaenus spumarius (L.) (Homoptera). tion. Climatic Significance and Environmental Impact (eds Hurrell Annales Zoologici Fennici, 4, 1–18. JW, Kushnir Y, Ottersen G, Visbeck M). Geophysical Monograph, Harper G, Whittaker JB (1976) The role of natural enemies in the 134, 235–262. colour polymorphism of Philaenus spumarius (L.). Journal of Ottersen G, Planque B, Belgrano A et al. (2001) Ecological effects Animal Ecology, 45, 91–104. of the North Atlantic Oscillation. Oecologia, 128, 1–14. Helle T, Kojola I, Timonen M (2001) Impact of snow cover on Post E, Forchammer MC (2002) Synchronization of animal the reindeer population in Ka¨sivarsi, NW Finland: is North population dynamics by large-scale climate. Nature, 420, Atlantic weather oscillation (NAO) involved? Suomen Riista, 168–171. 47, 75–85. (in Finnish). Post E, Stenseth NC (1998) Large-scale climatic fluctuation and Holmgren M, Scheffer M, Ezcurra E et al. (2001) El Nin˜o effects population dynamics of moose and white-tailed deer. Journal on the dynamics of terrestrial ecosystems. Trends in Ecology and of Animal Ecology, 67, 537–543. Evolution, 16, 89–94. Post E, Stenseth NC (1999) Climatic variability, plant phenology, Horsfield D (1978) Evidence for xylem feeding by Philaenus and northern ungulates. Ecology, 80, 1322–1339. spumarius (L.) (Homoptera: Cercopidae). Entomologia Experi- Qian B, Saunders MA (2003) Summer UK Temperature and Its mentalis et Applicata, 24, 95–99. Links to Preceding Eurasian Snow Cover, North Atlantic SSTs, Hunter MD, Price PW (1998) Delayed density-dependence and the NAO. Journal of Climate, 16, 4108–4120. or exogenous driving variables? Ecological Entomology, 23, Ranta E, Kaitala V, Lindstro¨m J (1999) Spatially autocorrelated 216–222. disturbances and patterns in population synchrony. Proceed- Hu¨ppop O, Hu¨ppop K (2003) North Atlantic Oscillation and ings of the Royal Society of London B, 266, 1851–1856. timing of spring migration in birds. Proceedings of the Royal Ranta E, Lundberg P, Kaitala V et al. (2000) Visibility of Society of London B, 270, 233–240. the environmental noise modulating population dynamics. Hurrell JW (1995) Decadal trends in the North Atlantic Oscilla- Proceedings of the Royal Society of London B, 267, 1851– tion: regional temperatures and precipitation. Science, 269, 1856. 676–679. Raven JA (1983) Phytophages of xylem and phloem: a compar- Hurrell JW, Kushnir Y, Ottersen G et al. (2003) An overview of the ison of animal and plant sap-feeders. Advances in Ecological North Atlantic Oscillation. In: The North Atlantic Oscillation. Research, 13, 135–234. r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2250–2262 2262 A. HALKKA et al.

Schaefer K, Denning AS, Leonard O (2004) The winter Arctic Tuomenvirta H, Drebs A, Frland E et al. (2001) Nordklim data Oscillation and the timing of snowmelt in Europe. Geophysical set 1.0 – description and illustrations. KLIMA Report 08/01, Research Letters, 31, L22205, doi: 10.1029/2004GL021035. Norwegian Meteorological Institute, Oslo, Norway, 26 pp. Silfverberg H (1968) On the distribution of Coleoptera Va¨ha¨talo AV, Rainio K, Lehikoinen A et al. (2004) Spring arrival in the Tva¨rminne archipelago. Notulae Entomologicae, 48, of birds depends on the North Atlantic Oscillation. Journal of 1–21. Avian Biology, 35, 210–216. Sparks YH, Roy DB, Dennis LH (2005) The influence of tem- Vandenbosch R (2003) Fluctuations of Vanessa cardui butterfly perature on migration of Lepidoptera into Britain. Global abundance with El Nin˜o and Pacific Decadal Oscillation cli- Change Biology, 11, 507–514. matic variables. Global Change Biology, 9, 785–790. Stenseth NC, Mysterud A (2005) Weather packages: finding the Weaver CR, King DR (1954) Meadow spittlebug Philaenus leu- right scale and composition for climate in ecology. Journal of cophthalmus (L.). Ohio Agricultural Experimental Station Research Animal Ecology, 74, 1195–1198. Bulletin, 741, 1–99. Stenseth NC, Mysterud A, Ottersen G et al. (2002) Ecological White TCR (2004) Limitation of populations by weather-driven effects of climate fluctuations. Science, 297, 1292–1296. changes in food: a challenge to density-dependent regulation. Stenseth NC, Ottersen G, Hurrell JW et al. (2003) Studying Oikos, 105, 664–666. climate effects on ecology through the use of climate indices: Whittaker JB (1969) The biology of (Diptera) para- the North Atlantic Oscillation, El Nin˜o Southern Oscillation sitising some British Cercopidae (Homoptera). Proceedings of and beyond. Proceedings of the Royal Society of London B, 270, the Royal Entomological Society, 44, 17–24. 2087–2096. Whittaker JB (1970) Cercopid spittle as a microhabitat. Oikos, 21, Stervander M, Lindstro¨mA˚ , Jonze´nNet al. (2005) Timing of 59–64. spring migration in birds: long-term trends, North Atlantic Whittaker JB (1973) Density regulation in a population of Philae- Oscillation and the significance of different migration routes. nus spumarius (L.) (Homoptera: Cercopidae). Journal of Animal Journal of Avian Biology, 36, 210–221. Ecology, 42, 163–172. Stiling P (1988) Density-dependent processes and key factors in Whittaker JB (2001) Insects and plants in a changing atmosphere. insect populations. Journal of Animal Ecology, 57, 581–593. Journal of Ecology, 89, 507–518.

r 2006 The Authors Journal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2250–2262