DISCUSSION

The fundamental aim of this investigation was to determine whether metapopulation theory could, as suggested by Hartvigsen & Halvorsen (1994), assist prediction of community composition and understanding of community heterogeneity. Although landscape ecology focuses on spatial heterogeneity it is equally applicable to temporal heterogeneity (Hartvigsen & Halvorsen, 1994) and their hypotheses can be tested on both scales. In their earlier study (Hartvigsen & Halvorsen, 1993) they compared parasite communities in brown trout in 4 lakes, 3 of which were unconnected. Indeed, most persons employing metapopulation theory seem to have envisaged that the local populations are not connected but are separated by unsuitable habitats (Hanski & Gilpin, 1991). This need not be the case and a lack of connection is not obligatory for application of the theory. There is therefore no reason why a river and its tributaries should not be viewed as a metapopulation and locations in the river and its tributaries as local populations. The only consequence of this is that one might predict higher than normal levels of similarity between communities given their proximity and connectivity (Poulin & Morand, 1999).

The choice of eels for testing the hypotheses was not simply due to pragmatic considerations such as ease of capture: they are particularly suitable because they are found in a wide range of aquatic conditions and are known to occur through the whole of the River Exe system. They are also easily able to move freely from one locality to another, unlike other species of fish which tend to be restricted to particular parts of the system. They will inevitably carry their parasites in such movements and so these have the potential to rescue declining local populations or replace extinct ones. Moreover, since all the helminths being studied are autogenic (Esch et al. 1988), this is the only way in which such rescue effects could operate. In addition, the helminth fauna of eels is well known, as is the extent of its variation in space (Kennedy, 1990; Kennedy & Guégan, 1996) and time (Kennedy, 1993a, 1997). Thus, both site and host species appear to be well suited to testing the hypotheses of (Hartvigsen & Halvorsen 1994).

The investigation focused on helminths in the intestinal habitat, since this harbours several species which form a recognizable community. Non-intestinal species were considered and analysed separately because they are not known to interact with any of the intestinal species and there was only a single species in each habitat. Moreover, A. crassus is known to be a recent introduction to the UK and to the River Exe catchment, throughout which it is still spreading rapidly, and the status of Pseudodactylogyrus sp. in the River Exe remains unclear (Kennedy, 1993b). These 2 species were included in the investigation for completeness but principally as further examples of patchy distributions, colonizing species and anthropochore influences on the parasite fauna. It seems inappropriate, however, to include them in spatial and temporal similarity comparisons of intestinal helminth communities to which they clearly do not belong.

The first prediction, that it is possible to recognize common and rare species, was undoubtedly met. Amongst the eel specialist intestinal helminths in the system, P. tenerrima, S. inermis and B. claviceps could be considered common and P. macrocephalus rare, whilst Pseudocapillaria sp. occupied an intermediate position. It is difficult to apply the core and satellite concept to these species since the method distinguishing these categories advocated by Hanski (1982a, b) is not readily applicable to species poor communities. Hartvigsen & Halvorsen (1993) used the concept in their study, but as descriptive terms only. If the same procedure is adopted here, both P. tenerrima and S. inermis could be described as core species. These core species, together with B. claviceps, were the most widespread and abundant species within the system, occurring in most localities and samples (in all for P. tenerrima) and one or other of them dominating all the communities except that in CU. Indeed these 3 species together are responsible for over 50% of the abundance in all communities except that in CU. This does appear to confirm the prediction that core species will be better at dispersal. This conclusion is, however, sensitive to scale as was also predicted. Although P. tenerrima is a widespread and common parasite of eels throughout Britain, it only dominates a minority of communities (Kennedy, 1990). By contrast, S. inermis has only previously been recorded from the British Isles in 3/87 localities. It is widespread but very local in continental Europe and is considered to be a rare species (Moravec, 1994). Why it is so common in the rivers Exe and Otter is not known: it was not found in the Otter before 1987 (Kennedy, 1997) or in the River Culm up to 1986 (Kennedy, unpublished), and it is possible that it has only recently colonized these systems. Another widespread species throughout Europe is A. clavula, which was once common and dominant in the eels of the River Clyst (Kennedy, 1984, 1993a) but can currently be considered uncommon. A different pattern again is shown by P. laevis, which can be locally very abundant as in the Culm and Otter but which is very patchily distributed on a regional scale (Kennedy, Bates & Brown, 1989). By contrast, Pseudocapillaria sp. is not widely distributed and always occurs in low abundance. The predictions relating to common and rare species do thus seem to be met on the whole, but the data do not completely support the view of Nee, Gregory & May (1991) that species abundant in one patch are also found in more patches.

The predictions relating to community similarity, by contrast, do not seem to be met. Each locality was connected to each of the others by water, a suitable habitat, and so connectivity of populations should be very high. However, similarity between communities bore no relationship to connectivity of localities. Indeed, the percentage similarity index ranged from 5% to 87%, with the majority of values below 25%. This range is similar to that reported by Esch et al. (1988) from eels in 12 localities scattered across Britain, whereas intuitively one would have expected similarity values to be far higher within a single river system. Similarity also appeared to be unrelated to proximity and distance between localities. Neighbouring localities showed both high (64%) and low (5%) levels of similarity and 3 of the highest similarity pairings involved a locality in the adjacent catchment. Moreover, average levels of similarity did not differ between localities in the main river and in the tributaries. The data thus refute this prediction of Hartvigsen & Halvorsen (1994) and do not support the views of Poulin & Morand (1999).

The predictions relating to change and turnover of species do appear at first to be met. One obvious source of spatial and temporal heterogeneity is the spread of an invading species throughout the system, exemplified by the natural spread of A. crassus and the so far inexplicable appearance of P. laevis in the River Culm some time post 1986 and its virtual restriction to that river. There is also evidence of species appearing in eel populations over the 3 years (C. lacustris), and of species disappearing from eel populations (N. rutili, E. truttae and P. macrocephalus), whilst others are erratic in respect of their appearance in, and disappearance from, particular localities. However, these changes do not per se provide evidence of a rescue effect or indeed of a disappearance from the system. Some of these species have a history of erratic appearances and disappearances from eel populations over much longer time scales in the River Clyst (Kennedy, 1993a) and the River Otter (Kennedy, 1997). Moreover, several of these species are specialists of other species of fish and their appearance or disappearance from eels and from any particular locality may reflect changes in, and movements of, the populations of their preferred hosts. The appearance of C. lacustris, for example, is likely to reflect changes in the perch population and changes in other species may reflect changes in the pike and trout populations and their movements throughout the system.

Perhaps, therefore, a test of the predictions should be confined to eel specialists only although the Culm helminth communities would then appear anomalous in being dominated by a non-specialist. The 3 common specialists P. tenerrima, S. inermis and B. claviceps show far greater population stability than the other species: between them they dominate all the samples, except those from the Culm; together they contribute most to the relative abundance of helminths in the communities, except for the Culm; and they are responsible for the levels of similarity between the communities and the changes therein. Changes in the identity and relative abundance of a dominant species in space or time are reflected in corresponding changes in levels of community similarity. The other two eel specialists, P. macrocephalus and Pseudocapillaria sp. always exhibited low abundance and so contributed little to community structure or similarity. There is no evidence either way for a rescue effect operating on these species and there appeared to be no place in the system where they were sufficiently abundant to serve as a source for a rescue. The only possible rescue effect is A. clavula. This species declined in the River Clyst after 1982, and a rescue effect from the River Exe was suggested for its continuation in the river (Kennedy, 1984). It has subsequently persisted in the Clyst (Kennedy, 1993a) and metapopulation theory still provides the most plausible explanation for the continuing local extinctions and re-appearances of this species.

Overall, however, metapopulation theory does not provide a very satisfactory explanation for the changes in helminth communities in eels in the River Exe system, as most of the predictions are not met in practice. There is evidence for species being common and rare, for heterogeneity in space and time and for change and turnover in populations but there are alternative and equally plausible explanations for most of these phenomena in terms of species invading the catchment or utilizing other fish species as principal hosts. They generally involve non-eel specialists or rare species, and can be equated with noise in the system. The ‘core' species of eels show great stability and no evidence of local extinctions or re-colonizations. They are the dominant species in the helminth communities of eels, except for the Culm, and their local and minor population changes do not provide any support for metapopulation theory.

It is pertinent to ask why this should be so. There are two, not necessarily exclusive, explanations. In the first place it is difficult to apply metapopulation theory to a particular species of parasite in a particular species of host, the efforts of Hartvigsen & Halvorsen (1994) and Simberloff & Moore (1997) notwithstanding. If the greater proportion of a parasite species cycles though another species of host, as is the case with C. truttae which uses brown trout and, P. laevis which cycles through chub, the changes in the helminth population and so community in eels will reflect events in the other host i.e. the true metapopulation is not being studied but only one small part of it. The definition of metapopulation used here is that of Hanski & Gilpin (1991) but not that of Simberloff & Moore (1997) who specifically define it as ‘all individuals of one parasite species in one population of one host species'. Such a definition may be appropriate for a monospecific parasite with a direct life-cycle, but it cannot apply to a parasite with wider specificity and employing intermediate hosts in its cycle. The definition of Hanski & Gilpin (1991) does allow for this. A consequence of this situation is that it may be necessary to study the parasite in all its hosts and not just one host, often a desirable but impracticable task. Alternatively, it is possible as suggested above to focus only on the specialist core species, but this also presents difficulties. Some species do not fit easily into the core-satellite mould, for example A. clavula which prefers eels as a definitive host but can infect a wide range of fish species. It would also rule out studies on all the communities such as those in the River Culm where the dominant ‘core' species is a non-specialist of eels. Parasite populations, and so subsequently communities, in the definitive host may also reflect changes in the intermediate host populations. The changes in A. clavula in the Clyst may, for example, reflect changes in the populations of its isopod intermediate host Asellus meridianus in relation to its competitor Asellus aquaticus (Kennedy, 1984). It is also possible that the restriction of P. laevis to the River Culm reflects the far higher densities of Gammarus pulex in this river than in other tributaries of the River Exe, with this in turn reflecting the higher levels of calcium and alkalinity in the River Culm compared to that of the River Exe (50/mgl and 126/mgl respectively compared to 15/mgl and 58/mg). Heterogeneity in distribution of invertebrate intermediate hosts must of necessity lead to heterogeneity in distribution of the adult population.

Secondly, the River Exe system is unusual in that eel helminth communities are dominated by eel specialists, apart from the Culm, and also in that S. inermis is a ‘core' species there. In the majority of British localities eel helminth communities are not dominated by eel specialists. In a survey of 50 localities, Kennedy (1990) found only 38.2% were dominated by an eel specialist and the single most common dominant species was Acanthocephalus lucii, the principal host of which is perch. A similar situation appears to prevail throughout continental Europe (Moravec, 1985; Saraiva & Eiras, 1996; Schabuss et al. 1997; Kennedy et al. 1998; Sures et al. 1999). As mentioned previously, S. inermis is considered a rare species (Moravec, 1985, 1994) and Kennedy (1990) recorded it dominating in only 2.1% of localities. Dominance of eel helminth communities by P. tenerrima and S. inermis, and indeed their status there as ‘core' species seems to be a peculiarity of the southwest of England and has been reported from no-where else. The River Exe system may thus be atypical. Hartvigsen & Halvorsen (1993) were perhaps fortunate in their choice of fish species and localities for the field study from which they developed their hypotheses. They studied the helminths of brown trout in small isolated lakes, of which 2 harboured no other species of fish, and all their communities were composed exclusively of salmonid specialists. This may well be a general feature of brown trout helminth communities compared to eel ones as 96.9% of 72 trout communities were dominated by salmonid specialists as opposed to 38.2% (48.8% if A. clavula is included) of eel ones (Kennedy & Hartvigsen, 2000). River Exe eels are more similar in this respect to Anguilla rostrata in Nova Scotia, where helminth communities are also dominated by eel specialists (Marcogliese & Cone, 1998). It may be, therefore, that the predictions of Hartvigsen & Halvorsen (1994) are only applicable under a restricted set of conditions. Eel helminth communities in the River Exe system do not meet these conditions, and here metapopulation theory does not assist understanding of the spatial and temporal changes in the intestinal helminth communities of eels.