Composition and ecological drivers of the kwongan scrub and woodlands in the northern Swan Coastal Plain, Western Australia

The nature of community patterns and environmental drivers in kwongan mediterranean-type shrubland on nutrient-poor soils occurring in Western Australia remain poorly examined. We aimed to determine whether (i) classification of the kwongan vegetation of the northern Swan Coastal Plain would be ecologically informative and (ii) which environmental drivers underpin the plant community patterns. The study area was positioned on the northern Swan Coastal Plain, locality of Cooljarloo (30°39′ S, 115°22′ E), situated 170 km north of Perth, Western Australia. Compositional (518 species × 337 relevés) and environmental data set (29 variables × 87 relevés) describing time since last fire, soil chemical and physical properties, and terrain characteristics were analysed using classification and ordination techniques. OptimClass assisted in the selection of a robust data transformation, resemblance function and clustering algorithm to identify the vegetation patterns. Major ecological drivers of the vegetation patterns were detected using distance-based redundancy analysis (db-RDA). Classification revealed major groupings of Wet Heath and Banksia Woodland distinguishable by the high prevalence of myrtyoid and proteoid taxa, respectively. On floristic-sociological grounds, we recognised four Wet Heath and two Banksia Woodland communities. The Wet Heath was constrained to areas of higher litter depth (db-RDA axis 1: 9%). Soil chemical and physical properties explained the highest proportion (17%) of the compositional variance, while the terrain- and fire-related variables explained 2% and <0.001%, respectively. While fire explained little compositional variance overall, a separate db-RDA analysis found that it may play an important pattern-structuring role within Banksia Woodlands. Fine-scale compositional patterns correspond only to a small extent to environmental data; the substantial unexplained variance may be due to slow-acting neutral and stochastic processes.