Soil baiting was used to monitor temporal and spatial soil inoculum dynamics of Phytophthora cinnamomi in disease centres in four Banksia woodland biomes of the Swan Coastal Plain bioregion and an Eucalyptus marginata forest biome of the Jarrah Forest bioregion of the South-West Botanical Province of Western Australia. Patterns of inoculum occurrence in P. cinnamomi disease centres in Banksia woodland biomes were similar to those in E. marginata forest biomes. Significantly lower frequency of isolation was from near-surface (top 0.03 m) soil and greater occurrence of viable inoculum from soil at depth (~1 m below the soil surface) and from near-surface soil around the collar of dead Banksia. Seasonal fluctuations in soil inoculum were not as distinct or consistent as that shown for rainfall and temperature and soil water content. Viable inoculum was most frequently isolated from near-surface soil in spring. In all Banksia woodland biomes isolation of inoculum from soil at depth was lowest in summer and greatest in autumn to spring. In all but one biome, isolation of inoculum from soil around dead Banksia was lowest in summer. Inoculum of P. cinnamomi was isolated from groundwater 3–5 m below the soil surface for Banksia woodland biomes in winter and spring 1987 and the winters of 1988 and 1989. Spatial patterns of viable inoculum from soil at depth were aggregated in 42% of quadrats analysed and 34% of quadrats sampled. The pattern of inoculum occurrence found in P. cinnamomi disease centres was influenced by creation of dynamic spatiotemporal niche refuges favourable to the pathogen through ecosystem engineering by host and pathogen.