Differential tree seedling recruitment across forest-savanna ecotones is poorly understood, but hypothesized to be influenced by vegetation cover and associated factors. In a 3-y-long field transplant experiment in the forest-savanna ecotone of Ghana, we assessed performance and root allocation of 864 seedlings for two forest (Khaya ivorensis and Terminalia superba) and two savanna (Khaya senegalensis and Terminalia macroptera) species in savanna woodland, closed-woodland and forest. Herbaceous vegetation biomass was significantly higher in savanna woodland (1.0 ± 0.4 kg m−2 vs 0.2 ± 0.1 kg m−2 in forest) and hence expected fire intensities, while some soil properties were improved in forest. Regardless, seedling survival declined significantly in the first-year dry-season for all species with huge declines for the forest species (50% vs 6% for Khaya and 16% vs 2% for Terminalia) by year 2. After 3 y, only savanna species survived in savanna woodland. However, best performance for savanna Khaya was in forest, but in savanna woodland for savanna Terminalia which also had the highest biomass fraction (0.8 ± 0.1 g g−1 vs 0.6 ± 0.1 g g−1 and 0.4 ± 0.1 g g−1) and starch concentration (27% ± 10% vs 15% ± 7% and 10% ± 4%) in roots relative to savanna and forest Khaya respectively. Our results demonstrate that tree cover variation has species-specific effects on tree seedling recruitment which is related to root storage functions.

In tropical West Africa, distribution patterns of forest islands in savannas are influenced by fires which occur regularly in the grass stratum. Along continuous forest–savanna transects in the Comoé National Park, the change in the amount and composition of non-woody phytomass was investigated from savanna to forest interior. This was correlated with the cover of vegetation strata above, soil depth, and the occurrence of seasonal surface fires. Phytomass mainly consisted of leaf litter in the forests (about 400 g m−2 at the end of the rainy season, and about 600 g m−2 at the end of the dry season) and of grasses in the savanna (about 900 g m−2). Low grass biomass appeared to be primarily the result of suppression by competing woody species and not of shallow soil. The occurrence of early dry-season fires seemed to be determined mainly by the amount of grass biomass as fuel because fires occurred in almost all savanna plots while forest sites remained unaffected. However, late dry-season fires will encounter higher amounts of leaf litter raising fire probability in forests. Due to the importance of the amount of combustible phytomass, fire probability and intensity might increase with annual precipitation in both savanna and forest.