EMPIRICAL TESTS OF SPECIFIC PREDICTIONS MADE FROM A SPATIAL MODEL OF THE POPULATION-DYNAMICS OF MACULINEA-REBELI, A PARASITIC BUTTERFLY OF RED ANT COLONIES

We test the ability of a complex spatial model to predict the size of M. rebeli populations in two mountain ranges. The published model was based on nineteen parameters that described the population dynamic int eractions between the butterfly, its initial food plant Gentiana cruci ata, its subsequent host Myrmica schencki, and other competing non-hos t species of Myrmica. All original model parameters were measured on o ne site in the Spanish Pyrenees. In this study we measured seven of th ese parameters, concerning gentians and their habitat, at 12 new sites in the French Alps and at another in Spain. We show that model predic tions using these values explained 86% of the variation in M. rebeli e gg numbers, independently measured on the other sites. Three other (no n-intuitive) model predictions were also partly confirmed: (i) A prior i, our model (which had been parameterised from a single, atypical sit e with very high gentian densities) predicted a close, indirect, non-l inear relationship between gentian and butterfly densities, with egg-n umbers being negatively correlated at high gentian densities and posit ively correlated over low densities. The prediction would be unusual f or phytophagous butterflies and is counter-intuitive for a species whi ch experiences only moderate competition on its food plant before spen ding most of its life inside ant nests, where it is regulated by stron g, density-dependent competition. Although this was not fully tested d ue to a lack of new sites supporting high gentian densities, the empir ical results showed a very close positive relationship between egg num bers and gentian densities on sites supporting < 1200 gentians per hec tare. Furthermore, our original mechanistic model also accurately pred icted both the slope and the range of the positive relationship found at low densities. (ii) Annual fluctuations in populations of M. rebeli are exceptionally small. The fit between observed and predicted egg d ensities was significantly closer than expected if M. rebeli populatio ns are assumed to fluctuate with the amplitude of the least variable p hytophagous British butterfly. (iii) Myrmica schencki populations on t he new sites occurred at the low densities predicted by the model as a consequence of parasitism by the butterfly. The accuracy of these pre dictions over a wide range of sites encourages us to apply the model t o simulate the effects of different conservation options for M. rebeli .