Habitat quality versus connectivity: Pool depth rivals isolation in determining the presence and cover of a dispersal-limited flora endemic to vernal pools

Background/Question/Methods

Successful restoration of native plant communities may depend as much on landscape spatial structure as on the maintenance of appropriate physical conditions at the restoration site.  We studied the relative influence of local and landscape variables using data from a ten-year experiment in restoring native flora to California vernal pools, ephemeral wetlands supporting many rare and endangered plant species.  Previously, we identified pool depth and seeding treatment as key predictors of local restoration success.  Here, we used multi-model inference to estimate the influence of pool depth and seeding treatment relative to the size and connectivity of experimental pools, and to identify the best model(s) of presence and cover for each of five focal species.  Using time-averaged data from unseeded control pools and from unseeded regions of seeded pools, we modeled the presence or cover of each species as a response to local pool depth, size, seeding intensity and/or selected metrics of connectivity, via generalized linear models.  Connectivity models were customized (in R) to fit parameters controlling effects of inter-pool distances, source pool sizes, and target pool size.  Models were ranked using an information criterion adjusted for sample size (AIC_c), and individual predictor variables were ranked across models using Akaike weights.

Results/Conclusions

We were able to fit models to presence and cover data for four of the five focal species.  Pool depth rivaled connectivity as the best predictor of presence and cover.  All species exhibited higher occupancy and cover in deeper, less isolated pools.  Pool depth and connectivity had positive effects in the top models for each species and for combined species cover, and bootstrapped 95% CIs for the coefficients of these predictors rarely overlapped zero.  Alternative connectivity models included effects of distances to all experimental pools, distances to all natural pools within the study site, distance to off-site vernal pools, and/or connectivity via waterways.  The best models of connectivity differed among species, but isolation rose rapidly with distance for every species.  We found little support for effects of seeding treatment, relative to other predictors of presence and cover, but some support for effects of target and source pool sizes on species presence, especially for the endangered species Lasthenia conjugens.  Surprisingly, our fitted models of connectivity, with estimated effects of target and source pool sizes and inter-pool distances, rarely improved on a rough classification of pools according to three classes of “distance” from natural pools in the center of the study site.