The effects of fertilization and herbivory on the herbaceous vegetation of the boreal forest in north-western Canada: a 10-year study

1 The influence of fertilizer addition and mammalian herbivore exclosures (a 2 x 2 factorial design, with four replicates at each of two sites) on the cover, species composition and diversity of the understorey vegetation of the boreal forest in the south-western Yukon, Canada, were investigated from 1990 to 1999. This was done to test whether vegetation composition was controlled by resource level alone (bottom-up control), herbivory alone (top-down control), or by both (interactive control). 2 The density of the major herbivore, the snowshoe hare, varied 25-fold, declining from 148 km(-2) in 1990 to 8 km(2) in 1994, and increasing to a second peak of 198 km(2) in 1998. 3 In control plots most species were remarkably constant in percent cover. After 10 years, most of the major species showed significant responses to fertilizer with four species increasing (Festuca altaica, Mertensia paniculata, Epilobium angustifolium, and Achillea millefolium), and three declining (Linnaea borealis, Lupinus arcticus, Arctostaphylos uva-ursi). Some species took up to 5 years before a response was detected. 4 Fertilization caused (i) a decline in the number of species, and species evenness in the community, (ii) a reduction in the proportion of woody species, and (iii) an increase in herbaceous dicotyledons and grasses. 5 The exclusion of herbivores had virtually no impact on the abundance of the vegetation or on species diversity, except in 1990-92 during a decline from a peak of 148 hares km (2) to 29 hares km(-2). 6 These results suggest that the percentage cover and composition of herbaceous vegetation in the boreal forest are determined almost exclusively by the productivity of the site (bottom-up control) and that the activities of mammalian herbivores may be important only during peaks in hare population densities (interactive control). 7 Results were both species-specific and time-dependent, suggesting that long-term studies are needed to discriminate between long-term responses to treatments and transient phenomena.