Tree-grass niche partitioning is pervasive and supported by water-use time series data
A large fraction of the terrestrial biosphere is comprised of ecosystems where trees and grasses coexist. Forecasting vegetation dynamics in these systems requires a thorough understanding of tree-grass interactions, including the extent to which trees and grasses partition soil moisture. In tropical savannas, there is conflicting evidence for niche partitioning across studies and locations. It is also still unclear whether snapshot data on tree-grass rooting differences (a preferred approach) actually imply ecologically meaningful differences in water use at seasonal scales. We used stable isotopes from soil and stem water to examine landscape-level patterns of tree-grass functional rooting differences across a rainfall gradient in a South African savanna ecosystem. We then conducted a targeted study to test the hypothesis that tree-grass functional rooting differences do translate into physiologically distinct responses to soil moisture availability as a function of depth. We measured sap-flow velocity in trees and canopy temperature (a transpiration correlate) in grasses, while tracking soil moisture content at multiple depths and meteorological variables over the course of a growing season. We used lasso regression to identify the dominant soil moisture layers driving daily variation in tree and grass water-use metrics while controlling for weather variables.
We found clear and pervasive differences in functional water uptake between grasses and trees based on the isotopic data, both at the landscape and local scales, with grasses clearly exhibiting shallow functional rooting profiles and trees being deeper-rooted. Our experimental approach showed that these patterns were strongly supported by the water-use data, which showed that grasses and trees predominantly responded to soil moisture availability at 5 and 40 cm depth, respectively. Our results suggest that rooting-depth differentiation between savanna trees and grasses is pervasive, and provide a rare example of mechanistic support for the resource partitioning hypothesis in the tropical savanna biome. These results have important implications for our understanding of tree-grass dynamics, including under altered precipitation regimes.