Résumé
Analyzing population dynamics in changing habitats is a prerequisite for
population dynamics forecasting. The recent development of metapopulation
modeling allows the estimation of dispersal kernels based on the
colonization pattern but the accuracy of these estimates compared with
direct estimates of the seed dispersal kernel has rarely been assessed. In
this study, we used recent genetic methods based on parentage analysis
(spatially explicit mating models) to estimate seed and pollen dispersal
kernels as well as seed and pollen immigration in fragmented urban
populations of the plant species Crepis sancta with contrasting patch
dynamics. Using two independent networks, we documented substantial seed
immigration and a highly restricted dispersal kernel. Moreover,
immigration heterogeneity among networks was consistent with previously
reported metapopulation dynamics, showing that colonization was mainly due
to external colonization in the first network (propagule rain) and local
colonization in the second network. We concluded that the differences in
urban patch dynamics are mainly due to seed immigration heterogeneity,
highlighting the importance of external population source in the
spatio-temporal dynamics of plants in a fragmented landscape. The results
show that indirect and direct methods were qualitatively consistent,
providing a proper interpretation of indirect estimates. This study
provides attempts to link genetic and demographic methods and show that
patch occupancy models may provide simple methods for analyzing population
dynamics in heterogeneous landscapes in the context of global change.