At Lamto, little is known about animal community responses to habitat variability resulting from fires and the mosaic pattern of the vegetation in general and in particular about that of termites which play key roles in this ecosystem. With a standardized method, data were collected on termites from four habitats differing in their vegetation cover and fire-history: annually burned savanna, savanna woodland, forest island and gallery forest. A range of environmental variables was measured and correlated with species abundances. The number of termite species collected in the savanna woodland was very close to that found in the gallery forest while the forest island was the richest habitat. The species richness of the savanna woodland and forest island seemed partly due to their heterogeneous and transitional vegetation structures and variable food resources. With regard to the fire-history of habitats, Connell's intermediate disturbance hypothesis offers an explanation for differences in the patterns of habitat-specific species richness. Variation in species abundances was significantly correlated with only two environmental variables (soil pH and woody plant species richness). The pH appeared as the most influential factor for fungus-growers while tree invasion in the savanna strongly reduces the abundance of grass-feeding species (e.g. Trinervitermes geminatus). Although not significantly correlated with species abundances, soil carbon showed a positive correlation with the dominant soil-feeder Basidentitermes potens. As for wood-feeders, they were not strongly correlated with woody plant species richness; this fact might be linked to their use for other sources of nourishment. Overall, it appears that habitat variability in the Lamto reserve contributes to the maintenance of different subsets of the termite community.

The position of Macrotermes michaelseni mounds were mapped in six areas of 1.5 to 31.0 ha in a range of habitats on the fringes of the Okavango delta. Abundance varied significantly between and within habitats with highest densities in relatively stable floodplain woodland. In all sites apparent overdispersion of active mounds was not statistically significant when edge effects were taken into account. When active and inactive mounds were combined some overdispersion was seen but due to the disturbance regime the probability of all mounds being active simultaneously was small. Intraspecific competition was, therefore, not considered likely. Our results and reanalyses of dispersion patterns in published accounts for Macrotermes species suggest that although mound dispersion can appear to be regular this is probably a result of historical precedence and chance events rather than any current density-dependent process.