Although there is a well‐known association between tree cover and soil texture in savannahs, the hydrological drivers of tree cover variation have not been systematically explored, particularly in parallel with factors such as fire, herbivory, and tree–grass interactions. The relationship between hydrological factors and tree cover is important for resolving the relative contribution of bottom‐up versus top‐down factors in structuring savannah vegetation. We quantified soil moisture dynamics across eight 1‐km transects spanning tree cover gradients from open to woody savannah in Serengeti National Park in Tanzania using soil moisture sensors coupled with dataloggers. We mapped tree cover at two spatial scales through supervised classification of high‐resolution satellite imagery. We simultaneously produced water retention curves in open and woody habitats within each transect to compare soil hydrological properties and to convert volumetric water content (θ) from dataloggers to plant‐available water over the course of an annual cycle. We also quantified grass biomass at 100 locations per transect, estimated fire frequency from MODIS satellite data, and quantified herbivore occupancy with paired camera traps situated in open and woody habitats within each transect. We found a positive relationship between tree cover and soil moisture drainage rate, and found that open habitats had more negative water potentials than woody habitats for a given value of θ. In contrast, we found no evidence for a consistent relationship between grass biomass or fire frequency and tree cover. We found evidence for higher browser occupancy in woody than open habitats, but no habitat effects on herbivores as a group (browsers plus grazers), suggesting that herbivory is unlikely to be the dominant factor explaining variation in tree cover. Synthesis. Our results suggest that variation in tree cover is partly driven by hydrological (edaphic) factors unrelated to fire, herbivory, tree–grass interactions or mean annual precipitation at these spatial scales in Serengeti. We contrast our findings with previous work attributing tree cover shifts in Serengeti to precipitation gradients.