Appendix A. A detailed escription of the conservation strategies used.
The conservation strategies used in our simulations are based on Oostermeijer et al. (1994, 1998), Maes et al. (2004 ), and G. Oostermeijer, personal communication. Note that the parameter ts measures the time since the beginning of G. pneumonanthe succession (ts = 0 means the conditions are optimal for G. pneumonanthe). We assumed that succession (from establishment to exclusion) occurs over 40 to 50 years (Chapman et al. 1989) . The effects of perturbations on parameters is explained below and listed in Table 3.
Sod cutting consists of removing the top layers of the soil, enabling colonization by G. pneumonanthe seeds. Cutting must be done carefully so as not to damage the habitat, and sufficiently close to G. pneumonanthe adults so that their seeds may germinate. Sod is generally cut in late autumn, when the positions of adult plants are clearly visible. It takes about 1 year before a sod-cut area is suitable for colonization by gentians, because a mildly toxic ammonia-peak occurs in the soil during the first year post-cut, which kills seedlings and juveniles. In conservation areas, it is generally recommended to cut several patches of sizes between 2 and 10 m2 (the total area depends on the size of the site, the person-power available, and the machinery). For large population complexes, the area to cut will be no more than 10% of the total area. Myrmica scabrinodis is also a quick colonizer of sod-cut areas, but special care must be taken to not damage too many of the Myrmica nests where M. alcon caterpillars are overwintering. The parameter ts representing succession age, is reduced according to the proportional area of soil removed. Because juvenile, vegetative adult, and dormant plants are too difficult to detect and protect, their survival will be reduced by sod cutting (percentage proportional to the percentage of area cut). The negative effect on Myrmica through the destruction of nests depends on the percentage of area removed. Similarly, the negative effect on M. alcon occurs through damage to ant nests.Mowing is frequently used to promote grassland G. pneumonanthe populations (Oostermeijer et al. 1994) . It must be done in late September when M. alcon larvae have already left G. pneumonanthe. The parameter ts will be reduced depending on the height of the mowing (between 5 and 20 cm). An important side-effect of mowing at 5 cm height will be the occasional creation (in a temporal and spatial sense) of small-scale sod-cuts that are made by slips of the mowing machine (i.e., locally, ts is set back to 0). The effect of mowing on G. pneumonanthe will be positive through the reduction of succession age ts. A negative impact on G. pneumonanthe survival (juveniles, vegetative, and reproductive adults) occurs only if mowing is done at 5cm or less. Moreover, juvenile survival is reduced somewhat by decomposing cuttings and direct mortality caused by the mowing machinery, but adult plants tolerate this well. Mowing is favorable to Myrmica scabrinodis because it reduces ts. Colony size will also be affected negatively by mowing, because higher/larger nest mounds cannot be created in short-mown sites and the Myrmica tend to have smaller, more widely dispersed nests. In our model, this will result in a reduced probability for M. alcon caterpillars to be found by ants after leaving G. pneumonanthe, and a smaller number of caterpillars that can be supported per ant nest.
Grazing by cattle, sheep, horses, and occasionally deer creates short turf, and their hoof-prints expose bare soil. This is advantageous to G. pneumonanthe (Oostermeijer et al. 1994) . Ideally, periodic grazing should occur after late September to avoid negative effects on M. alcon eggs and larvae, or on G. pneumonanthe seed production. In practice, it is difficult to restrict grazing on nature reserves, and low-intensity year-round (extensive) grazing is a more ‘natural’ way of keeping the vegetation structure open (G. Oostermeijer, personal communication). When periodic grazing occurs, the succession age ts will be reduced, depending on the intensity of grazing. In the case of year-round grazing it will tend to keep ts at a relatively stable level, depending on the intensity of grazing. The survival of G. pneumonanthe adults will not be strongly affected, but some juveniles will be killed by intensive trampling (depending on stocking levels). Flowering/seed production will also be affected, depending on grazing intensity. The effect of grazing on Myrmica is positive via the reduction (and increased stability) of ts. The direct effect on M. alcon will be negative in the case of year-round grazing through the death of eggs and larvae (depending on the intensity of grazing). This negative effect is particularly strong in small populations of G. pneumonanthe where nearly all plants may be grazed.
LITERATURE CITED
Chapman, S. B., and R. J. Rose. 1982. Pages 74–78 in Ecological studies on the marsh gentian (Gentiana pneumonanthe). Institute of Terrestrial Ecology. Annual Report for 1982.
Chapman, S. B., R. J. Rose, and R. T. Clarke. 1989. The behavior of populations of the marsh Gentian (Gentiana pneumonanthe) - a modeling approach. Journal of Applied Ecology 26:1059–1072.
Maes, D., H. van Dyck, W. Vanreusel, and W. Talloen. 2004. Functional conservation units for the endangered butterfly Maculinea alcon in Belgium (Lepidoptera: Lycaenidae). Biological Conservation 120:233–245.
Oostermeijer, J. G. B., S. H. Luijten, Z. V. Krenova, and H. C. M. Den Nijs. 1998. Relationships between population and habitat characteristics and reproduction of the rare Gentiana pneumonanthe L. Conservation Biology 12:1042–1053.
Oostermeijer, J. G. B., R. Vantveer, and J. C. M. Dennijs. 1994. Population-structure of the rare, long-lived perennial Gentiana pneumonanthe in relation to vegetation and management in the Netherlands. Journal of Applied Ecology 31:428–438.