Abstract
QuestionWhat are the relationships between edaphic factors, fire regime and woody vegetation structure and composition in savannas at landscape to regional scales?LocationThe Transboundary ‘W’ Regional Park, Niger, West Africa.MethodsWe focused on a protected savanna ecosystem with rainfall of 700 mm yr−1, where rainfall is not expected to be a strong limiting factor for tree cover, and with historically low grazer and browser densities. A burned area history map was created over 7 yr using the high temporal resolution Aqua MODIS space-borne sensor. In the field, the composition and structural parameters of the woody layer, as well as soil samples, were acquired in 137 plots of 0.09 ha each, based on a stratified random sampling approach to sampling the fire regime (seasonality and frequency). Using classical multivariate methods, we analysed the correlations between woody cover characteristics, fire regime and soil and geomorphological factors.ResultsIn spite of management practices aimed at generalizing early fire every year, the fire regime created a diverse spatial structure, with some vegetation less prone to burning (fire-free) and other types that burned very frequently. These diverse fire histories showed correlations with gradients of both structure and composition of the woody layer. Soil variables had a higher explanatory power for vegetation structure and composition than fire.ConclusionsAmong the plausible causal chains between fire regime, soil factors and woody vegetation structure and composition, our results showed better agreement with a model in which fire regime is more a consequence than a cause of the other factors. This contrasts with bottleneck/perturbation theories as well as management practices, but meets the common sense opinions of field practitioners. Although our regional approach should be complemented by controlled experiments at a local scale, it does appear that investments in annual efforts to organize early fire campaigns in many protected areas may need to be reassessed.