P. Giraudoux, Department of Environmental Biology, EA3184 usc INRA, University of Franche-Comté, Place Leclerc, F-25030 Besançon, Cedex, France (fax +33 03 81 66 57 97; e-mail firstname.lastname@example.org ).
1Outbreaks of the water vole Arvicola terrestris cause severe damage in grasslands of upland regions of Europe. The sheer speed of this phenomenon is a challenge to effective pest control measures. While there has been some research into factors that promote outbreaks, especially landscape composition, little is yet known of the biological mechanisms underlying the speed of colonization of grasslands during the population growth stage. Like A. terrestris, the mole Talpa europaea digs vast tunnel networks that may be used by A. terrestris. The availability of extensive networks of this sort might greatly boost the colonization potential of A. terrestris and so explain the speed of onset of its outbreaks.
2Areas occupied by networks of A. terrestris and T. europaea tunnels, and their respective locations, were evaluated and mapped with a global positioning system (GPS) during low-density, growth and the first weeks of abundance phases of cyclic fluctuations of A. terrestris.
3During the growth phase and the first weeks of the abundance phase nearly 80% of new A. terrestris colonies were found in T. europaea tunnel networks, although these occupied just 20% of the area under study.
4The regulating influence of wooded areas on the risk of outbreaks, which can be identified at larger spatial scales (region, sector), was imperceptible at local scales, probably because of the influence and relative extent of T. europaea tunnel networks.
5There was a threshold level of occupancy of grassland by T. europaea, as indexed by surface indices, below which A. terrestris populations were regulated naturally and over long periods.
6Synthesis and applications. These findings have implications for controlling outbreaks of A. terrestris. Temporally, chemical pest control of A. terrestris can be reduced by taking action during the A. terrestris low-density phases, preferably in the autumn. Spatially, control operations should be targeted at T. europaea and early A. terrestris networks. Mechanical destruction of tunnels (e.g. ploughing) and trapping should be considered as an alternative to chemical pest control.
In various regions of France and Switzerland populations of the water vole Arvicola terrestris L. fluctuate widely (Saucy 1988; Giraudoux et al. 1997). Their cycles last some 6–7 years and go through phases of very low density alternating with outbreak phases (Morel & Meylan 1970; Habert 1988; Saucy 1994). Outbreaks last 2–4 years and cause extensive damage in the grassland ecosystems of several upland regions (Habert, Pascal & Pradier 1991; Quéréet al. 1999). This phenomenon is affected by various factors relating to landscape composition (connectivity of grassland, extent of woodlots, openness of the environment; Giraudoux et al. 1997; Duhamel et al. 2000; Fichet et al. 2000). However, little is understood of other factors affecting the population cycle. In particular, the population growth phase is often astonishingly fast and commonly completed in a single breeding season (April–November). During this period, population densities fluctuate locally from a few individuals per hectare in the spring to several hundred individuals per hectare by late autumn.
The extent of networks tunnelled by the mole Talpa europaea L. during the phases of low population density of A. terrestris might account for the speed of this phenomenon. Talpa europaea and A. terrestris often share or replace one another in the same tunnel networks. The availability of extensive tunnels for populations of A. terrestris would mean the energy they saved on digging and reconfiguring tunnels could be expended on breeding. Reduced energy expenditure would also reduce the death rate of individuals during the dispersion phase. Few studies to date have looked at the interaction between these two species despite the presumed high level of competition between them, especially for space (Fritschy 1979; Meylan & Höhn 1991; Giraudoux et al. 1995).
Understanding the factors influencing the establishment of the first colonies of A. terrestris and accelerating their speed of colonization would provide a basis on which to develop strategies to control this species. As a contribution to this objective, this study tested the hypothesis that by digging or maintaining tunnels during low population density phases of A. terrestris, T. europaea promotes the establishment and faster expansion of populations of A. terrestris during the growth phase of its population cycle.
Testing this hypothesis first entailed answering the question: do gallery networks of T. europea affect the establishment and speed of growth of A. terrestris colonies? If the answer is affirmative, in an endeavour to quantify this process we evaluated the threshold level of T. europaea tunnels above which outbreaks occur much more quickly.
Ten study sites were selected among four different landscape types in the canton of Nozeroy in Franche-Comté, France (6·09°E, 46·8°N; see Appendix S1 in the Supplementary material). Landscape types were defined by their structure (number of kilometres of hedgerows) and openness (proportion of woodlots), measured within a radius of 250 m around the sites: (i) loosely structured (LS), very open landscapes or ‘open fields’ (less than 1000 m of hedgerows and less than 1 ha of woodland within the 250-m radius); (ii) moderately structured (MS), open landscapes (1000–2000 m of hedgerows and 1–2 ha of woodland); (iii) tightly structured by hedgerows (TSH), landscape criss-crossed by hedgerows (3000–5000 m of hedgerows and less than 1 ha of woodland); (iv) tightly structured by woodlots (TSW), woodland mosaic (less than 3000 m of hedgerows and 5–15 ha of woodland). Each study site was chosen to be in the centre of areas of these four landscape types while avoiding the immediate vicinity of villages known to impact negatively A. terrestris demographics (Delattre et al. 1996). At each site from two to seven habitat patches were selected, with a total of 32 patches for all 10 sites. Patch areas varied from 0·2 to 1·1 ha with a total patch area of 17 ha.
a. terrestrisandt. europaeacounts
Several sampling methods based on capture–mark–recapture and catch per unit effort are available (Pascal & Meylan 1986) and are usually statistically unbiased. However, such methods can only be applied over restricted spatial areas (1–2 ha) because they are demanding in terms of logistics, and they are limited in terms of study duration (1–2 months) because of the disturbance to individuals and populations from intensive trapping. To overcome these problems, relative abundance can be evaluated by using activity indices. Surface indices of A. terrestris and T. europaea activity (earth mounds and mole hills) are quite conspicuous. Giraudoux et al. (1995) give the main distinguishing features as tumulus shape (flat-topped or conical), proximity (touching or isolated mounds), spatial pattern (clustered or linear) and the presence or absence of ‘earth sausages’ (sausage-shaped clumps of earth found in fresh mole hills).
Arvicola terrestris and T. europaea indices were sampled simultaneously on each patch. The geographical coordinates of each A. terrestris colony and each T. europaea tunnel network were determined by a global positioning system (GPS) (GeoExplorer® 3, Fidd Data Solutions Inc., Jerome 83338, Idaho, USA). Networks of deep tunnels only, as identified by multiple criteria including the alignment (for T. europaea) and clustering (for A. terrestris) of earth mounds, were recorded (Giraudoux et al. 1995). A geographical information system (GIS) database was collated using ArcView. Indications of presence (earth mounds) are only approximate indications of the location of the actual network. A 2·5-m buffer zone was therefore defined around each object. It was used to determine and measure the zones where A. terrestris colonies developed in existing T. europaea networks.
In April 2002, when the A. terrestris populations had been in the low-density phase for several months, the area of each patch was measured and the initial level of abundance of the A. terrestris and T. europaea populations was evaluated. Five surveys were then made, two during the low-density phase of A. terrestris (May and June 2002), two during its growth phase (September and December 2002) and one at the beginning of its abundance phase (June 2003). Each time, the area of newly dug T. europaea networks was evaluated and new A. terrestris colonies were located and categorized as being either inside or outside the T. europaea networks.
Because of the longevity of T. europaea networks (more than 1 year for the deep tunnels identified) and because of the sampling method employed, network size was dependent on network size in the previous month. Because of this temporal dependency samples were not independent; this was accounted for in the test design below. Statistics and graphical display were computed with R (R Development Core Team 2004). Four questions were addressed.
Do colonies evolve preferentially inside T. europaea networks?
Under the null hypothesis of a random distribution of new colonies of A. terrestris, the expected proportion of A. terrestris new colonies within T. europea networks equals the proportion of T. europea network in a parcel. The null hypothesis was tested by a chi-squared test.
Was the formation of new colonies dependent on the area covered by T. europaea networks?
We examined the effect of patch area and T. europea network size upon the number of new colonies formed using generalized linear models with a Poisson error distribution. This was a natural model choice in which the expected number of new colonies varied in relation to patch size. Three variations of this model were considered. The first model contained the effect of patch area, the second model the effects of patch area and network size, and the third model the effects of patch area, network size and the interaction between those two variables. Models were compared using the information theoretic approach, as outlined by Burnham & Anderson (2002), and presented according to Anderson et al. (2001).
Were there any landscape-related effects?
As patch size in each landscape type did not follow a normal distribution and could not be transformed to Gaussian, a Kruskal–Wallis test was used for comparisons of ranks between categories, followed by a multiple comparison test as necessary, according to Siegel & Castellan (1984).
Do levels of T. europaea colonization impact the colonization dynamics of A. terrestris?
Visual analysis of the data suggested population growth was exponential during the study period; however, because measurements were repeated on patches a mixed effects model was required to account for non-independence between measurements at each patch. The percentage of patch area colonized by A. terrestris was modelled as an exponential function of the form:
where AT is the percentage area of patches occupied by A. terrestris, TIME is measured in months (integers) and a and b are model parameters with patch treated as a random effect on parameter a according to Pinheiro & Bates (2000). Three alternative parameterizations for b were investigated. In the first model, b was treated as a constant. In the second, b was modelled on T. europaea occupancy category. In the third, b was modelled on T. europaea occupancy category and a time–T. europaea interaction. As above, models were compared using the information theoretic approach. If variation in colonization rates between patches was not sufficiently explained by the model then this would be indicated by visualization of model residuals and a random effect for parameter b might be required. However, this was not indicated in the current analysis.
In April 2002, when A. terrestris populations were in the low density phase, T. europaea networks were locally extensive (Fig. 1). Patch occupancy by T. europaea networks ranged from 0% to 32%, with an overall mean of 14%, whereas figures for A. terrestris networks ranged from 0% to 27%, with an overall mean of just 3%. Over this period the highest levels of A. terrestris colonization were observed in patches where T. europaea networks were scarcest. From April to May 2002 the overall abundance of A. terrestris populations declined before increasing constantly and rapidly until December (= growth phase). An abundance phase was then established and continued beyond June 2003.
formation of newa. terrestriscolonies
Success in the formation of new A. terrestris colonies was measured by answering three questions.
During the growth phase did colonies evolve preferentially inside T. europaea networks, or outside those networks in a random manner?
If A. terrestris formed new colonies independently of the T. europaea networks, the proportion of new colonies inside and outside those networks should simply reflect the proportion of areas with and without T. europaea. Measurements showed that new colonies appeared mostly inside T. europaea networks, with 84% (May) and 74% (December) of new colonies appearing in T. europaea networks, although these networks covered only a little more than 20% of the total suitable area under study. In May, June, September and December 2002 any random distribution of A. terrestris colonies could be ruled out, their numbers being considerably overrepresented in T. europaea networks (Table 1).
Table 1. Number of new colonies of A. terrestris and their settlement inside (AT in) or outside (AT out) T. europaea networks. Ho, proportion of T. europaea network area in patches; χ2, chi-square value; d.f., degree of freedom; P, chi-square probability
Number of new colonies
Expected proportion (Ho)
As patches were not all the same size, were the formation of new colonies, the area covered by T. europaea networks and the area of each patch independent variables?
In May, June and September, Akaike information criterion (AICc) and Akaike weights were lower and higher, respectively, when models included T. europea networks. In December there was no evidence of an effect of T. europaea (Table 2).
Table 2. Comparison of models (response = number of new A. terrestris colonies) with and without T. europea network terms. PA, patch area; TE, T. europea network area; a, b, c regression coefficients; LL, maximized log-likelihood; K, number of estimated parameters; n/K, number of observations/K; AICc, second-order Akaike index criterion; Δi, difference between AICc and the lowest value of AICc; wi, Akaike weights
a.PA + b.TE
a.PA + b.TE + c.PA.TE
a.PA + b.TE
a.PA + b.TE + c.PA.TE
a.PA + b.TE
a.PA + b.TE + c.PA.TE
a.PA + b.TE
a.PA + b.TE + c.PA.TE
As the patches were distributed across different environmental contexts, was there any landscape-related effect?
Median patch sizes for landscapes with little structuring (MS and LS), landscapes structured mostly by woodlots (TSW) and landscapes tightly structured by hedgerows (TSH) were 0·43, 0·47, 0·37 and 0·77 ha, respectively. Patches in TSH were significantly larger than in TSW landscape types (Kruskal–Wallis 12·8, P= 0·005, multiple comparison test P < 0·05).
In all cases, a Kruskal–Wallis test failed to reject the null hypothesis of no difference in the proportion of patches occupied by T. europaea according to landscape types.
evaluation of a threshold level of t. europaeacolonization above whicha. terrestriscolonization may be promoted
In making this evaluation, all of the sample patches were subdivided into three subsamples corresponding to three T. europaea occupancy classes of similar frequencies measured in April 2002: (i) class A, 0–10% of the patch area occupied by T. europaea networks (mean 5·2%); (ii) class B, 10–20% (mean 15·3%); (iii) class C, 20–30% (mean 25·0%). These three classes comprised, respectively, 8, 13 and 11 patches of average areas of 4800, 5600 and 5500 m2.
The model corresponding to exponential growth with time and including T. europaea and T. europaea–time interaction effects minimized the AICc (Table 3). Figure 2 shows the mean fitted values from model 3 calculated according to the level of T. europaea occupancy. The model suggested a link between growth speed and the level of T. europaea colonization as assessed at the start of the study. The patch proportion occupied by A. terrestris was on average 1·5 times lower by the end of the study period in parcels where the level of T. europaea colonization was low (class A) compared to patches where the level of T. europaea colonization was higher (classes B and C) at the start of the study.
Table 3. Comparison of models (response = patch proportion occupied by A. terrestris). LL, maximized log-likelihood; K, number of estimated parameters; n/K, number of observations/K; AICc, second-order Akaike index criterion; Δi, difference between AICc and minimum AICc; wi, Akaike weights; CAT, category of T. europaea occupancy; TIME, time (months), α, β, γ, regression parameters on model parameter b (see the Methods)
α + β.CAT
α + β.CAT + γ.CAT.TIME
t. europaea–a. terrestrisinteraction
The extension of grassland areas in upland regions (Alps, Jura, Massif Central, Pyrenees) since the 1950s has been associated with a simultaneous increase in the amplitude of population fluctuations of T. europaea and A. terrestris. The apparently close interaction between the two species is evident in the alternating abundance peaks and their respective cyclic fluctuations (Giraudoux et al. 1995). Populations of T. europaea reach maximum abundance when populations of A. terrestris are in the low-density phase and vice versa. Observations by Fritschy (1979) suggest that A. terrestris populations tend to occupy T. europaea tunnel networks rather than excavate their own. As the availability of T. europaea networks is at its maximum at the end of the A. terrestris low population density phase, it is at this time that the T. europaea–A. terrestris interaction should logically prove most beneficial to the development of A. terrestris populations.
Our findings confirm that A. terrestris populations benefit from availability of T. europea tunnel networks, and that A. terrestris populations recolonize patches largely by exploiting existing T. europaea tunnels. In May 2002, during the low population density phase of A. terrestris, the distribution of its colonies was largely dependent on T. europaea tunnel networks. In June, colonization was already well underway. It varied both with the extent of T. europaea networks and with patch size. On large patches A. terrestris populations were more numerous and the number of new colonies greater. In December, A. terrestris dynamics appeared to be independent of T. europaea networks.
spatio-temporal pattern of interactions betweent. europaea anda. terrestris
The period during which T. europaea networks seem to play an important role is ultimately very short. It extends from April to May, although it may also include the preceding winter months when climatic conditions (frozen ground and snow cover) preclude any reliable observations. Spatially, the phenomenon is localized, being restricted to patches where T. europaea networks are present. Allowance has also to be made for patch size. The larger patches contain more new colonies. For a given patch size the T. europaea population size determines the number of new colonies of A. terrestris.
One non-exclusive explanation of these findings may well lie in the various forms of predation. The facility with which A. terrestris colonies grow and prosper in large T. europaea networks is probably associated with the difficulties predators have in capturing A. terrestris in such situations. Our protocol did not systematically include counts of digging by foxes in A. terrestris colonies. However, it was apparent that patches with non-extensive T. europaea networks were visited much more by predators and that from one sampling session to the next the rate of disappearance of A. terrestris colonies outside T. europaea networks was very high in these locations (P. Delattre & R. Clarac, unpublished data).
Apparently, T. europaea networks have so great an effect on A. terrestris populations during this period of the A. terrestris cycle that they completely override the effects of landscape, which are known to be important when examined over the entire cycle (Giraudoux et al. 1997; Duhamel et al. 2000; Fichet et al. 2000). Indeed, these authors have shown that in the Jura mountains and the Massif Central, landscape structures studied and compared on larger scales (regional and sectorial) regularly determine the sources of population outbreaks and their pattern of spread from one area to another during successive cycles. Over the short-term and at a local scale, on which most pest control strategies are commonly devised, the influence of T. europaea networks seems more important than that of landscape structures and must therefore be considered as one of the biggest risk factors.
Talpa europaea influences the demography of A. terrestris both during the settlement phase of A. terrestris (by enabling or facilitating the establishment of new colonies) and during the patch colonization phase (by accelerating colonization). By making allowance for the interaction between T. europaea and A. terrestris it should be possible to control A. terrestris populations at low population densities and very locally. Chemical pest control would then be just one of a range of options. It might also be useful to reconsider periodic trapping campaigns (Varlet & Lacombe 1961; Rudge 1963). Realistically this could only be feasible over sufficiently large areas given sufficient human resources, and would therefore need to be organized through umbrella organizations such as farmer collectives. The destruction of tunnels of all kinds by mechanical processes should be undertaken from the end of the outbreak. Where chemical pest control is considered to be the most effective option, products with fewer side effects should be used in preference to anti-coagulants (e.g. hydrogen phosphides, which are not transferred to small mammal predators in the food chain). Over the long-term, measures to protect predator species, restructuring of open environments and changes in fodder crop production (Giraudoux et al. 1997) provide a helpful framework for avoiding environmentally harmful practices, which are incompatible with the sustainable management of grassland ecosystems.
The authors are grateful to Nicolas Ehrhardt for his participation in the GPS surveys. Financial support was provided by the Regional Council of Franche-Comté and the regional delegation of the Ministry of Environment. They also thank the three anonymous referees and the editor for their most constructive comments on the early versions of the manuscript.