The quantities and spatial distribution of nutrients in savanna ecosystems are affected by many factors, of which fire, herbivory and symbiotic N 2fixation are particularly important. We measured soil nitrogen (N) pools and the relative abundance of N and phosphorus (P) in herbaceous vegetation in five vegetation types in a humid savanna in Tanzania. We also performed a factorial fertilization experiment to investigate which nutrients most limit herbaceous production. N pools in the top 10 cm of soil were low at sites where fires were frequent, and higher in areas with woody legume encroachment, or high herbivore excretion. Biomass production was co-limited by N and P at sites that were frequently burnt or heavily grazed by native herbivores. In contrast, aboveground production was limited by N in areas receiving large amounts of excreta from livestock. N 2 -fixation by woody legumes did not lead to P-limitation, but did increase the availability of N relative to P. We conclude that the effects of fire, herbivory and N 2fixation upon soil N pools and N:P-stoichiometry in savanna ecosystems are, to a large extent, predictable.Electronic supplementary material: The online version of this article (

We enriched in CO2 the canopy of 14 broad-leaved trees in a species-rich, ca. 30-m-tall forest in NW Switzerland to test whether elevated CO2 reduces water use in mature forest trees. Measurements of sap flux density (JS) were made prior to CO2 enrichment (summer 2000) and throughout the first whole growing season of CO2 exposure (2001) using the constant heat-flow technique. The short-term responses of sap flux to brief (1.5-3 h) interruptions of CO2 enrichment were also examined. There were no significant a priori differences in morphological and physiological traits between trees which were later exposed to elevated CO2 (n=14) and trees later used as controls (n=19). Over the entire growing season, CO2 enrichment resulted in an average 10.7% reduction in mean daily JS across all species compared to control trees. Responses were most pronounced in Carpinus, Acer, Prunus and Tilia, smaller in Quercus and close to zero in Fagus trees. The JS of treated trees significantly increased by 7% upon transient exposure to ambient CO2 concentrations at noon. Hence, responses of the different species were, in the short term, similar in magnitude to those observed over the whole season (though opposite because of the reversed treatment). The reductions in mean JS of CO2-enriched trees were high (22%) under conditions of low evaporative demand (vapour pressure deficit, VPD <5 hPa) and small (2%) when mean daily VPD was greater than 10 hPa. During a relatively dry period, the effect of elevated CO2 on JS even appeared to be reversed. These results suggest that daily water savings by CO2-enriched trees may have accumulated to a significantly improved water status by the time when control trees were short of soil moisture. Our data indicate that the magnitude of CO2 effects on stand transpiration will depend on rainfall regimes and the relative abundance of the different species, being more pronounced under humid conditions and in stands dominated by species such as Carpinus and negligible in mono-specific Fagus forests.

Availabilities of nitrogen (N) and phosphorus (P) have a strong influence on plant growth and the species composition of savannas, but it is not clear how these availabilities depend on factors such as fire, N 2-fixation, and activities of wild herbivores and cattle. We quantified soil N and P availabilities in various ways (extractable pools, mineralization, resin adsorption) along vegetation gradients within a recently abandoned cattle ranch and a former game reserve in Tanzania (both areas now part of the Saadani National Park). We also assessed annual N and P balances to evaluate how long-term availabilities of N and P are affected by large herbivores, symbiotic N 2-fixation, and fire. The results show that cattle ranching led to a spatial redistribution of nutrients, with the local accumulation of P being stronger and more persistent than that of N. In the former game reserve, intensively grazed patches of short grass tended to have elevated soil N and P availabilities; however, because quantities of nutrients removed through grazing exceeded returns in dung and urine, the nutrient balances of these patches were negative. In dense Acacia stands, N 2-fixation increased N availability and caused a net annual N input. Fire was the major cause for nutrient losses from tallgrass savanna, and estimated N inputs from the atmosphere and symbiotic N 2-fixation were insufficient to compensate for these losses. Our results call into question the common assumption that N budgets in annually burned savanna are balanced; rather, these ecosystems are a mosaic of patches with both N enrichment and impoverishment, which vary according to the vegetation type.