Correspondence author. 208 Mueller Building, University Park, PA16802, USA; E-mail: firstname.lastname@example.org
1Masting events in the autumn provide a carbohydrate-rich pulse of resources that can influence the dynamics of small mammals and their natural enemies. Similar patterns are observed with the periodical cicada emergence which provides a protein-rich pulse in the spring, but comparisons are confounded by timing and food type.
2We compared the influence of a naturally occurring spring pulse of cicadas with an experimental spring pulse of carbohydrate-rich seeds. We used a replicated population level field experiment and capture–mark–recapture techniques to record the vital rates, demographics, and abundance of Peromyscus leucopus (the white-footed mouse), as well as other small mammals and their parasites.
3The density of P. leucopus on grids where cicadas emerged was 55% higher than controls as a consequence of early breeding. This was followed by an increase in the prevalence of the nematode Pterygodermatities peromysci, reduced breeding and decreased recruitment rates. Other small mammals including Tamias striatus (eastern chipmunk) and Blarina brevicauda (short-tailed shrew), increased in density, but there was no affect on Sorex cinereus (masked shrew).
4In contrast to the presence of cicadas, there was no influence of sunflower seed supplementation on small mammal density, vital rates, or reproduction with the exception of an increase in B. brevicauda density. The response of small mammals to seasonal pulses depends on timing, food type, and species.
Seasonal breeding is a dominant pattern in nature and invariably coincides with temporal or spatial variation in resource abundance (Asdell 1964). Further to this, unpredictable pulses (those that do not occur annually) can perturb the system dramatically and have long-term consequences for the dynamics and the community structure. For example, unpredictable pulses in the fall, like mast seed production, may influence dispersal and survival of mice whereas spring pulses are more likely to influence breeding. The theory of resource enrichment has identified a paradox whereby resource enrichment may have no influence on abundance of the consumer when the benefits are taken up by natural enemies, but may also destabilize the relationship between consumer and natural enemy and even lead to extinction (Rosenzweig 1971). However, empirical work has rarely been able to support these predictions and much of the research has focused on the tensions between the stabilizing and destabilizing forces (Roy & Chattopadhyay 2007). Identifying and understanding the consequences of unpredictable enrichment and the nature of these pulses (protein or carbohydrate) is interesting, sometimes non-intuitive and has important ramifications for pest management, biocontrol, and conservation (Ostfeld et al. 1998; Ostfeld & Keesing 2000).
In a recent paper, Marcello, Wilder & Meikle (2008) examined if the unpredictable spring pulse of perishable cicadas had the same effect on P. leucopus as an autumnal pulse of cacheable acorns. They found good evidence that the spring emergence resulted in increased density similar to that observed in the fall (Wolff 1996; McShea 2000; Yunger 2002; Elias, Witham & Hunter 2004) but unlike the fall, this increase was a consequence of earlier and more breeding. Observational studies that focus on such complex effects are naturally confounded in time, space and community structure and designing good comparative controls can be difficult. Carbohydrate pulses provide an energy-rich food source for overwinter survival, whereas protein-rich cicadas may provide the limiting amino acids needed for breeding. P. leucopus and Blarina brevicauda (Say) have been shown to opportunistically eat these insects, but the relative responses are expected to vary between insectivorous and omnivorous species (Hahus & Smith 1990). A longitudinal study that examined the relationship between cicada emergence and small mammal abundance identified a fourfold increase in shrew (B. brevicauda) abundance but no response by P. leucopus (Krohne, Couillard & Riddle 1991).
We studied the 2004 emergence of Brood X cicadas (also studied by Marcello, Wilder & Meikle 2008), a brood being composed of three species: Magicada septendecim (Fisher), M. cassini (Linnaeus), and M. septendecula (Alexander & Moore) that emerges synchronously after 17 years. They are one of 20 cicada broods in eastern North America that emerged in an area stretching from Princeton (New Jersey) south to Baltimore (Maryland) and west into Ohio across southern Pennsylvania. To examine the effects of pulsed food type on the dynamics of small mammals, we compared control populations to those with added sunflower seeds (carbohydrate pulse) and no cicadas to those with cicadas present (protein rich). Response variables included small mammal density, weekly recruitment rates and vital rates, and the response by parasites as a proxy for natural enemies. We predicted an increase in density, breeding, immigration and parasites in areas with supplemental food or cicadas but expected a stronger response in the protein-rich cicada grids.
Materials and methods
experimental design and trapping
Live trapping of small mammals was used to monitor relative abundance before, during, and after the cicada emergence. The majority of cicadas in the area emerged on the morning of 3 June 2004 and we trapped until 19 August 2004. We established nine 8 × 8 grids of multi-capture live traps (Ugglan Special, Graham, Sweden) with 15-m intervals between traps and each grid at least 250 m between grids. Traps were checked for two consecutive days each week until 7 July and then bi-weekly until mid-August. To check for any longer-term effects on abundance or differences between sites, we returned to the control and cicada grids 2 years later in late May and trapped both control and cicada grids with snap traps. All grids were in typical, mature open hardwood forest typical of the north-eastern Appalachians where the forest is dominated by Quercus spp. and Acer spp. Three grids were located in Huntingdon County, Pennsylvania, where the periodical cicadas emerged. The other six grids (three controls and three sunflower seed addition) were in Centre County, Pennsylvania, just north of the emergence where cicadas were not present (Fig. 1). Three of these grids were randomly assigned to receive supplemental sunflower seeds (18 kg week−1 grid−1) for the first three weeks of trapping and this corresponded with the time period when cicadas were present on the three grids in Huntingdon County. During the fourth week when the cicadas died and fell to the ground, an additional 45 kg of sunflower seeds were spread by hand on each of these three supplemental food grids to mimic this additional pulse.
vital rate measures, density, and dissections for macroparasites
Standard body measures were taken and recorded for each capture of P. leucopus and Tamias striatus (Linnaeus). These measures included body length, tail length, body mass, sex, body condition, and breeding status. Animals were considered in breeding condition if they had descended testes (males), or if they were lactating, pregnant, or had a perforate vagina (females). Individuals were identified using a Trovan™ (Electronic ID Devices, Ltd., Santa Barbara, CA, USA) passive-induced transponder (PIT) tag inserted into the scruff of each animal. Individuals caught in more than one trapping session were considered residents. Population size estimates for P. leucopus and T. striatus were estimated using Jolly–Seber (see Pollock et al. 1990). The abundance estimates for the Sorex cinereus (Kerr) and B. brevicauda are transformed with natural logarithms [ln(x + 1)]; to avoid stress and conform to Institutional Animal Care and Use Committee (IACUC) requirements, these animals were not marked but released immediately on capture. All P. leucopus caught during the last trapping session (n = 150) were euthanized with an overdose of isofluorane and the intestinal tracts dissected and all gastrointestinal worms collected, identified and counted. This experiment was conducted with the approval of the Pennsylvania State Animal Care Committee (IACUC #16061, ‘Transmission Dynamics of Directly Transmitted Diseases in Wildlife Reservoir Hosts’).
Statistical analyses were conducted with r (http://www.r-project.org). Nested analyses of variance (anova) (error term = grid) were used to analyse response variables such as body mass, body condition, days known alive, recruitment, and time in breeding condition, in order to ensure that grid was the unit of replication. We obtained estimates of recruitment for each grid using a Pradel model (Pradel 1996) in the program mark (White & Burnham 1999). Abundance data and the Jolly–Seber population size estimates were analysed with nested anova with week as a covariate. When the response variable was binomial, the proportions were arcsine square root-transformed and analysed with Gaussian linear models with week as a covariate. The best predictors of each response variable were selected using backwards stepwise selection, retaining variables with P < 0·05 based on F-statistics. When significant differences were detected in the full model, we divided the data set to see which treatment effects were significant.
We undertook 6912 trap nights and caught 1527 small mammals: 827 P. leucopus, 241 T. striatus, 247 B. brevicauda, 182 S. cinereus, 24 Myodes gapperi (= Clethrionomys gapperi; Boreal Redback Vole; Vigors), 3 Zapus hudsonius (Meadow Jumping Mouse; Zimmermann) and 3 Glaucomys volans (Southern Flying Squirrel; Linnaeus). A lack of abundance of the last three species prevented any further analysis.
p. leucopus abundance
Abundance of P. leucopus was not altered by the addition of sunflower seeds [coef. = −0·097, F = 0·00, d.f. = (1, 4), P = 0·98] but, as predicted, was significantly greater on the grids with cicadas as compared to the controls [coef. = −5·21, F = 6·85, d.f. = (1, 4), one-tailed P = 0·029; Fig. 2a]. The difference in abundance between the sunflower addition grids and the cicada grids was not significant [coef. = −5·31, F = 1·81, d.f. = (1, 4), P = 0·25] (Fig. 2a). Jolly–Seber methods do not permit an estimate of population size for the first trapping period, before cicada emergence, so we used the log-transformed number of animals caught as the response variable and there was no significant difference between control and cicada grids during this week of trapping [coef. = 7·66, F = 4·44, d.f. = (1, 4), P = 0·10]. We repeated this analysis for the period while cicadas were present (sessions 2, 3, and 4) and compared with the subsequent period when cicadas were absent (sessions 5 and 6) and found significantly more P. leucopus when cicadas were present [coef. = 10·78, F = 61·10, d.f. = (1, 4), P = 0·0014] but not after they had gone [coef. = −2·17, F = 1·57, d.f. = (1, 4), P = 0·28]. We also returned to the control and cicada sites 2 years after the study and found no significant difference in the density index of P. leucopus between sites [coef. = 1·06, F = 0·039, d.f. = (1, 4), P = 0·86].
p. leucopus vital rates
First, we present results for P. leucopus vital rates, and then breeding, recruitment, survival, and macroparasite abundance. Subsequently, we report findings for T. striatus and B. brevicauda. There was no significant effect of treatment (either cicada presence or the additional sunflower seeds) on body mass, body length, or body condition when we examined all animals and considered sex (all P > 0·05).
A Gaussian linear model identified a significantly lower proportion of animals in breeding condition on the cicada grids than either the fed (coef. = −0·26, t = −6·73, P = 0·002) or the not fed control grids (coef. = −0·23, t = −14·72, P = 0·0001) but no significant difference between sunflower fed and control grids (coef. = −0·015, t = −0·341, P = 0·75; Fig. 3a). This pattern held for both males (coef. = −0·32, t = –3·54, P = 0·024) and females (coef. = −2·04, t = −3·01, P = 0·04) when we compare the control and cicada-supplemented populations. When we evaluated the proportion breeding between sunflower seed fed grids and control grids, we found no significant differences for males (coef. = −0·10, t = −0·90, P = 0·42) or females (coef. = −0·048, t = 0·25, P = 0·81), nor were there significant differences in the proportion breeding between fed and cicada grids for either males (coef. = –2·04, t = −2·35, P = 0·078) or females (coef. = −0·25, t = −1·34, P = 0·24; Fig. 3b–d).
To examine breeding effort in more detail, we analysed the proportion of time individual residents were in breeding condition. In the full model, the effect of treatment was highly significant [coef. fed = 0·52, coef. not fed = 0·47, F = 35·71, d.f. = (2, 6), P = 0·0004] and this pattern held for both males [coef. fed = 0·43, coef. not fed = 0·45, F = 5·87, d.f. = (2, 6), P = 0·04] and females [coef. fed = 0·72, coef. not fed = 0·45, F = 13·99, d.f. = (2, 6), P = 0·005]. The animals on both the sunflower fed [coef. = 0·44, F = 115·15, d.f. = (2, 6), P = 0·0004] and control [coef. = 0·45, F = 89·17, d.f. = (2, 6), P = 0·0007] grids were in breeding condition for a significantly greater proportion of time than animals on grids where the cicadas emerged. There was no significant difference between sunflower fed and not fed control grids [coef. = 0·004, F = 0·04, d.f. = (2, 6), P = 0·95; Fig. 3d]. The results of our survival models indicated no significant differences associated with either treatment [coef. = 2·78, F = 1·39, d.f. = (1, 4) P = 0·47; Fig. 3e].
The Pradel model estimates of recruitment indicated the effect of treatment was close to being statistically significant, but counter to what was expected. The control populations had higher recruitment than the sites supplemented with cicadas [coef. = 0·44, d.f. = (1, 4), F = 6·75, P = 0·06] or sunflower seeds [coef. = 0·47, d.f. = (1, 4), F = 4·92, P = 0·09]. The recruitment onto cicada grids did not differ from the fed grids [coef. = 0·66, d.f. = (2, 6), F = 1·27, P = 0·32; Fig. 4a]. Since MARK does not permit examination of the recruitment of individual cohorts of animals, we used the log-transformed number of captures through time of the three mass classes of mice (juveniles < 16 g, sub-adults 16–19·9 g, and adults > 20 g) and estimated recruitment rate as the slope of the relationship between number of new recruits and time for each grid. Models with these recruitment rates as the response variables identified the effects of treatment were highly significant for juveniles [coef. = 0·44, 0·54, F = 34·59, d.f. = (2, 6), P = 0·0005], approached significance for sub-adults [coef. = 0·075, 0·18, F = 3·77, d.f. = (2, 6), P = 0·087] and were not significant for adults [coef. = 0·086, 0·10, F = 3·77, d.f. = (2, 6), P = 0·57; Fig. 4b]. The recruitment of juveniles onto cicada grids was initially high but fell to be lower than control grids, whereas juvenile recruitment increased on these control grids (Fig. 4c).
The survival of resident animals was not significantly altered by either the addition of sunflower seeds or the presence of cicadas [coef. = 0·27, F = 0·85, d.f. = (1, 4), P = 0·41] and this was true for both males [coef. = 0·19, F = 0·33, d.f. = (1, 4), P = 0·60] and females [coef. = 0·38, F = 0·85, d.f. = (1, 4), P = 0·40; Fig. 3f].
p. leucopus natural enemies
We evaluated the intestinal parasite intensity of the 150 P. leucopus captured and euthanized on the last trapping session. Four nematode species: Pterygodermatities peromysci (Lichtenfels), Capillaria americana (Read), Syphacia peromysci (Harkema), and Heligmosomoides vandegrifti (Durette-Desset & Kinsella); as well as one cestode species (Hymenolepis s. str.) and one trematode species (Brachylaima peromysci, Reynolds) were recorded. The most common helminth (P. peromysci) was twice as prevalent (52%) on the cicada grids when compared to the fed (24%) and not fed (19%) grids and the effect of treatment was significant [χ2 = 13·7, d.f. = (2, 147), P = 0·001]. There was a significant interaction between sex and treatment which only marginally lowered the Akaike information criterion (AIC) of the model (177·15 to 175·02). When sexes were examined separately, the influence of treatment on P. peromysci prevalence was only significant for male mice [χ2 = 18·8, d.f. = (2, 77), P < 0·001], not females [χ2 = 1·9, d.f. = (2, 56), P = 0·39]. The difference in prevalence between the fed and not fed grids was not significant for either males [χ2 = 0·02, d.f. = (1, 53), P = 0·90] or females [χ2 = 1·1, d.f. = (1, 35), P = 0·28]. The pattern of higher prevalence of P. peromysci on the cicada grids holds for the comparisons between both fed [χ2 = 9·4, d.f. = (1, 94), P = 0·002] and not fed [χ2 = 10·8, d.f. = (1, 101), P = 0·001] grids and in both of these comparisons, we find the prevalence was only significant in males [fed: χ2 = 13·2, d.f. = (1, 46), P = 0·0003; not fed: χ2 = 14·6, d.f. = (1, 55), P = 0·001] and not females [fed: χ2 = 0·01, d.f. = (1, 37), P = 0·87; not fed: χ2 = 1·7, d.f. = (1,40), P = 0·20; Fig. 4d].
eastern chipmunk: tamias striatus
The predicted effect of treatment on the abundance of T. striatus was significant when we considered all treatments [coef. = −1·39, –4·37, F = 4·74, d.f. = (2, 6), one-tailed P = 0·029]; with no statistical difference between either sunflower-fed and not-fed grids [coef. = −2·98, F = 4·44, d.f. = (1, 4), P = 0·10] or between cicada grids and the controls [coef. = −1·39, F = 1·67, d.f. = (1, 4), one-tailed P = 0·13]. There was a tendency (P < 0·1 > 0·05) towards greater abundance in the cicada grids when compared with fed grids [coef. = −0·44, F = 6·07, d.f. = (1, 4), P = 0·069] (Fig. 2b). We again used the log-transformed capture data for each grid to circumvent the lack of a population size estimate for the first trapping session and found the abundance of T. striatus on cicada grids was not different than the control during the first trapping session [coef. = 0·33, F = 0·1, d.f. = (1, 4), P = 0·76] or the final two trapping sessions [coef. = 3·00, F = 4·70, d.f. = (1, 4), P = 0·096] but was higher on cicada grids while the cicadas were present [coef. = 3·56, F = 14·62, d.f. = (1, 4), P = 0·019].
shrew species: b. brevicauda and s. cinereus
The insectivores (B. brevicauda and S. cinereus) were not PIT tagged and so capture–mark–recapture techniques could not be used to estimate abundance. We examined log abundance of animals caught per grid per trapping session. The effect of treatment was highly significant on the abundance of both B. brevicauda [coef. = −1·71, –1·13, F = 87·0, d.f. = (2, 6), P < 0·0001] and S. cinereus [coef. = 0·97, 1·04, F = 16·3, d.f. = (2, 6), P < 0·003]. The abundance of B. brevicauda was highest on cicada grids and this was significantly greater than both control [coef. = −1·70, F = 139·6, d.f. = (1, 4), P = 0·0003] and fed grids [coef. = −1·13, F = 141·26, d.f. = (1, 4), P = 0·0002]. Interestingly, this insectivore was also significantly more abundant on the sunflower seed-fed grids than the not-fed controls [coef. = 0·57, F = 14·92, d.f. = (1, 4), P = 0·018] (Fig. 2d). S. cinereus showed no such effect between fed and not-fed grids [coef. = 0·079, F = 0·15, d.f. = (1, 4), P = 0·72] but had significantly higher abundance on both fed [coef. = 1·05, F = 47·72, d.f. = (1, 4), P = 0·002] and not-fed grids [coef. = 0·968, F = 15·459, d.f. = (1, 4), P = 0·017] than on the cicada grids (Fig. 2c).
We examined the hypothesis that the abundance and vital rates of small mammals would be increased by a springtime pulse in resource availability as per Marcello et al. (2008). Specifically, we tested between the effects of protein-rich cicadas and carbohydrate-rich sunflower seeds and predicted that both should increase the vital rates and abundance of the small mammal community, but we postulated this influence would be greater on the grids with cicadas. We predicted a larger influence of cicadas not only because the timing of the pulse coincides with the beginning of the breeding season, but also because the higher protein content would provide amino acids which have been shown to be limiting nutrients in the springtime (McAdam & Millar 1999a,b). We found that the springtime emergence of cicadas led to a brief increase in abundance of P. leucopus, T. striatus, and B. brevicauda but not S. cinereus. This appeared to be caused by reproduction before the cicada emerged and was followed by reduced breeding and reduced immigration. In contrast, on the sunflower seed grids, the abundance of small mammals did not increase, with the exception of the shrew B. brevicauda. Overall, the spring response of P. leucopus to cicadas and seed was different to the fall seed pulses; the response to cicadas was short lived and led to a decrease in breeding and an increase in parasitism. We conclude from this that the effects of a protein-rich pulse do indeed have different effects on small mammal dynamics than a carbohydrate-rich pulse in spring.
In the study by Marcello et al. (2008), the demographic process giving rise to increased density was higher breeding production associated with earlier breeding, before cicada emergence. We provide further evidence to suggest that increased breeding occurred earlier on grids where cicadas emerged in that we found a high rate of juvenile recruitment during the first two weeks of the emergence. It is interesting to note that overall, there was a tendency (P = 0·06) after the emergence had started for more animals to be recruited onto the control grids than either the cicada or fed grids and this was counter to our a priori predictions based on observations from fall pulses (Fig. 4a). The analyses of the temporal patterns of recruitment found the recruitment rate onto cicada grids declined during the study (Fig. 4b). Given these findings, we suppose the increased density observed was a consequence of improved breeding before the emergence (Fig. 4b). Analyses of juvenile recruitment onto cicada grids shows a high initial recruitment which was then followed by a decline, and this contrasts to what was observed on the control and carbohydrate-supplemented grids (Fig. 4c).
During the period of cicada emergence, there was reduced breeding among the adults with a smaller proportion of the animals in breeding condition and those that were breeding, did not remain in breeding condition for as long. P. leucopus exhibits a midsummer breeding hiatus when they cease breeding for a period of several weeks in midsummer (Burt 1940; Wolff 1986; Terman 1998). The evidence from this study coupled with that provided by Marcello et al. (2008) suggests that the mice commenced breeding on the cicada grids earlier than usual and as such entered the summer breeding hiatus earlier than normal. Interestingly, the P. leucopus on the study sites with cicadas also carried greater intensities of P. peromysci, suggesting that these animals may have entered their summer breeding hiatus early because of the increased levels of parasitism. Pedersen & Grieves (2008) found food and parasitism interact to cause a late summer decrease in abundance and Vandegrift, Raffel & Hudson (2008) showed that the midsummer breeding hiatus can be reduced through the removal of gastrointestinal worms, notably the nematode P. peromysci.
The design of these pulsed resource experiments is confounded by a number of variables that need to be addressed. We expected the cicadas to emerge further north than they actually did and indeed as they have in previous years (Cooley, Marshall & Simon 2004), so grids we established to measure initial density and breeding effort had to be abandoned. This meant that the grids further south had only one trapping session just as the emergence began and this not only prevented us from using the Jolly–Seber techniques to estimate relative abundance before the pulse but also prohibited us recording in detail the higher reproductive output before the emergence (Marcello et al. 2008). Strictly speaking, the data are spatially confounded, since the sites where the cicadas emerged could not be selected at random, so the differences observed could be a reflection of site rather than treatment. We attempted to ameliorate this problem by first analyzing the data with respect to when the cicadas were present and second, by returning 2 years after the effects of the cicadas had passed so we could estimate density again. Together these data support the explanation that the significant differences in density were associated with the cicada emergence rather than the site or other effects.
A key finding in this study is that the addition of carbohydrate-rich sunflower seeds in the spring had no effects on density of P. leucopus or T. striatus compared with the controls. We suspect that this is related to differences in the protein and carbohydrate levels but we must also note that the response of animals to experimental fall pulses (as described in the introduction) is highly variable and it would have been informative to have a comparative fall experiment. Interestingly, B. brevicauda did increase in response to the addition of sunflower seeds although there is no clear explanation for why this occurred unless the seeds were attracting an invertebrate food for the shrew.
In summary, we found the pulse of protein-rich cicadas led to earlier breeding and high levels of juvenile recruitment and thus increased abundance. This was quickly followed by an increase in P. peromysci prevalence, a cessation of breeding, and a return to control levels of abundance. In contrast, the sites with supplemental sunflower seeds were indistinguishable from the controls. Interestingly, the springtime pulse led to a rapid increase in parasites which appeared to reduce breeding in a manner expected from the paradox of enrichment (Rosenzweig 1971). The springtime pulse was similar to the fall pulse in its duration, but the population responses were short lived in contrast to the masting events (Wolff 1996; Withham, & Hunter Elias et al. 2004; Falls et al. 2007). Albeit, we show a protein pulse provides sufficient nutrients to jumpstart springtime breeding, but this is counteracted by the effects of parasitism. In contrast, the high carbohydrate supplement had little influence and these patterns now need to be tested in the fall.
We wish to thank The Black Moshannon State Forest and the PA state Game Commission for permitting our field research. J.A. Sinclair, K.L. Krichbaum, L.K. Giebel, B.O. Ormond, B.D. Baylor, and A.D. Luis helped with field work. J. M. Kinsella has been a tremendous asset in his identification of gut macroparasites. We also thank the reviewers for their valuable comments. This project was funded by NSF grant #'s 0520468 and 0516227.