Understanding the relative influence of environmental and spatial variables in driving variation in species diversity and composition is an important and growing area of ecological research. We examined how fire, local vegetation structure and landscape configuration interact to influence dung beetle communities in Amazonian savannas, using both hierarchical partitioning and variance partitioning techniques to quantify independent effects. We captured a total of 3,334 dung beetles from 15 species at 22 savanna plots in 2003. The species accumulation curve was close to reaching an asymptote at a regional scale, but curves were variable at the plot level where total abundance ranged from 17 to 410 individuals. Most plots were dominated by just three species that accounted for 87.7% of all individuals sampled. Hierarchical partitioning revealed the strong independent and positive effect of percentage forest cover in the surrounding landscape on total dung beetle abundance and species richness, and richness of uncommon species and the tunneler guild. Forest cover also had a negative effect on community evenness. None of the variables that related to fire affected community metrics. The minimal direct influence of fire was supported by variance partitioning: partialling out the influence of spatial position and vegetation removed all of the individual explanation attributable to fire, whereas 8% of the variance was explained by vegetation and 28% was explained by spatial variables (when partialling out effects of the other two variables). Space-fire and vegetation-fire joint effects explained 14 and 10% of the dung beetle community variability, respectively. These results suggest that much of the variation in dung beetle assemblages in savannas can be attributed to the spatial location of sites, forest cover (which increased the occurrence of uncommon species), and the indirect effects of fires on vegetation (that was also dependent on spatial location). Our study also highlights the utility of partitioning techniques for examining the importance of environment variables such as fire that can be strongly influenced by spatial location.