156
S. Harrison
t = 2.16, P = 0.030; Helianthus, t = 2.65, P = 0.008; indicating these two species
had higher rates of colonization than the others), and showed a marginally significant decrease with increasing isolation (t = -1.91, P = 0.057; Fig. 2); the overall
model was again significant (2 log likelihood = 20.97, 6 df, P = 0.002).
We also censused population densities for three of the seep species at 50 sites
across a 20 x 40 km region, in 1997 and 1998. Between these years, Helianthus
exilis decreased strongly, Delphinium uliginosum increased moderately, and Mimulus
nudatus increased strongly at virtually every site in the region. Population fluctuations within each species were so completely synchronous at a regional scale that
neither population density, nor change in density from 1997 to 1998, was significantly spatially autocorrelated using standard analyses (Cliff and Ord 1981; Isaaks
and Srivastava 1989).
This study demonstrates that metapopulation dynamics may be relevant to the
persistence of rare plants in patchy environments (Menges 1990). Like many plants,
the species we studied are subject to regional-scale fluctuations driven by weather,
and probably rely in part on long-lived dormant seeds to survive in their harsh and
fluctuating environment. These considerations would appear to argue against the
importance of metapopulation processes. Nonetheless, we found a detectable role
for spatial isolation in determining the likelihood of local extinction and
recolonization (cf. Sjogren Gulve 1994), implying an important role for large-scale
dispersal in population persistence.
7 Discussion
The goal of this study was to examine as many facets as possible of how natural
habitat discontinuity affects population persistence and community structure in an
empirical system. While the study was guided in a general way by theoretical ideas,
the intention was not to choose a single theory or hypothesis and then study the
species most likely to exemplify it. Instead the study attempted to weigh the relevance of a variety of spatial theories to this particular study system.
Results of the diversity study were relatively supportive of the island biogeographic prediction of lower local diversity on islands compared with mainlands
(MacArthur and Wilson 1967). But these results also went beyond island biogeography by considering total (regional) and differentiation (beta), as well as local
(alpha) diversity. However, the results did not agree well with competition-based
metacommunity models predicting strong effects of patchiness on regional diversity (Caswell and Cohen 1991, 1993; Tilman et al. 1994). Instead they were strikingly consistent with the prediction (Holt 1997) of a higher ratio of habitat generalists to specialists in rare or patchy communities. Holt's prediction was based in part
on the idea that patches are influenced by the influx of propagules from the surrounding matrix (Wiens 1997). The results on edge effects in alien grasses supported this idea. However, while most spatial models treat extinction as a random
S. Harrison
t = 2.16, P = 0.030; Helianthus, t = 2.65, P = 0.008; indicating these two species
had higher rates of colonization than the others), and showed a marginally significant decrease with increasing isolation (t = -1.91, P = 0.057; Fig. 2); the overall
model was again significant (2 log likelihood = 20.97, 6 df, P = 0.002).
We also censused population densities for three of the seep species at 50 sites
across a 20 x 40 km region, in 1997 and 1998. Between these years, Helianthus
exilis decreased strongly, Delphinium uliginosum increased moderately, and Mimulus
nudatus increased strongly at virtually every site in the region. Population fluctuations within each species were so completely synchronous at a regional scale that
neither population density, nor change in density from 1997 to 1998, was significantly spatially autocorrelated using standard analyses (Cliff and Ord 1981; Isaaks
and Srivastava 1989).
This study demonstrates that metapopulation dynamics may be relevant to the
persistence of rare plants in patchy environments (Menges 1990). Like many plants,
the species we studied are subject to regional-scale fluctuations driven by weather,
and probably rely in part on long-lived dormant seeds to survive in their harsh and
fluctuating environment. These considerations would appear to argue against the
importance of metapopulation processes. Nonetheless, we found a detectable role
for spatial isolation in determining the likelihood of local extinction and
recolonization (cf. Sjogren Gulve 1994), implying an important role for large-scale
dispersal in population persistence.
7 Discussion
The goal of this study was to examine as many facets as possible of how natural
habitat discontinuity affects population persistence and community structure in an
empirical system. While the study was guided in a general way by theoretical ideas,
the intention was not to choose a single theory or hypothesis and then study the
species most likely to exemplify it. Instead the study attempted to weigh the relevance of a variety of spatial theories to this particular study system.
Results of the diversity study were relatively supportive of the island biogeographic prediction of lower local diversity on islands compared with mainlands
(MacArthur and Wilson 1967). But these results also went beyond island biogeography by considering total (regional) and differentiation (beta), as well as local
(alpha) diversity. However, the results did not agree well with competition-based
metacommunity models predicting strong effects of patchiness on regional diversity (Caswell and Cohen 1991, 1993; Tilman et al. 1994). Instead they were strikingly consistent with the prediction (Holt 1997) of a higher ratio of habitat generalists to specialists in rare or patchy communities. Holt's prediction was based in part
on the idea that patches are influenced by the influx of propagules from the surrounding matrix (Wiens 1997). The results on edge effects in alien grasses supported this idea. However, while most spatial models treat extinction as a random