There is a current interest in understanding how species abundances and distributions relate to environmental factors: species distribution models are one approach to quantifying the relationship between species and their environments. Models of species abundance, rather than distribution, may provide additional information regarding habitat suitability and species reproductive success and vulnerability to extinction, making abundance useful for conservation and land-management. Predictive modelling approaches can be problematic, however, given the possibility of multiple collinearities among explanatory variables, and the difficulty of identifying ecologically relevant factors. As well, species distribution models are assumed to be less accurate for systems in which disturbances are present but not explicitly modelled. For example, SDMs rarely account for fire-regimes, despite the acknowledged role of fires in determining the likelihood of colonization and persistence among species. Statistical methods such as variance partitioning and hierarchical partitioning focus on understanding the importance of ecological factors rather than on developing predictive models. Using these approaches, this study examines the relative contribution of fire, climate, and space to variation in the abundance of 28 Proteaceae species found in the fire-dominated Cape Floristic Region of South Africa.  

Results/Conclusions

On average, 71% of the deviance in species abundance can be explained using the combination of space, climate, and fire. The majority of variation was explained by the shared effects of fire and climate, suggesting that fire intensity and frequency are strongly collinear with climatic variables: of the independently-explained variance, space had the highest explanatory power, with fire and climate explaining relatively little independent variation. Comparisons between resprouting and reseeding species (two species that demonstrate contrasting adaptations to fire) indicated that resprouting species have, on average, greater unexplained variance and lower redundancy between fire and climatic variables. Hierarchical partitioning results highlighted that the relative importance of soil and fire-related variables tended to be species-specific, but that variables relating to seasonal precipitation and evapotranspiration, minimum temperature, and species abundance in neighbouring cells tended to be collectively important. The relatively low contribution of many variables, and the differences in variance partitioning among resprouting and reseeding species highlights the importance of incorporating life-history information when selecting variables for predictive models.