Variable distributions of water used for transpiration can have consequences on estimates of plant and ecosystem transpiration, as well as on the production potential and stability of transpiration rates for different plants within an ecosystem. Here a suite of studies, from the level of an individual tree stem to an ecosystem, illustrates this. At the tree level, the distribution of water flow within tree stems with tracheid xylem anatomy and deep sapwood was mathematically described. The distribution was Gaussian shaped, with a higher proportion of water flow in the outer sapwood compared to inner sapwood. The steepness of this distribution, , decreased over time with increasing leaf-to-air vapor pressure deficit and decreasing soil moisture availability. At the stand level, the proportion of nighttime stored water as a source for daytime transpiration increased with seasonal declines in soil moisture availability. I modeled the difference between potential stand canopy transpiration under no soil moisture deficit compared to the actual canopy transpiration observed when soil moisture declined. Had the trees not experienced declining soil moisture, canopy transpiration would have been 12 mm higher over the 25-day drought period. This specific modeling approach illustrates how multiple driving variables can influence transpiration on different time scales, and in linear and nonlinear ways. Finally, at the ecosystem level, I investigated the processes by which the dominant savanna plant life forms, trees and grasses, partition and use water, and how these feedback onto carbon gain. Across a natural hydrologic gradient, we found that trees primarily used groundwater, while grasses used a higher proportion of soil water. This had ramifications for aboveground plant production and variability in transpiration rates over time, with variability in tree transpiration rates always being less than that of the grasses. For the grass plant functional type, I found that the percentage of groundwater utilized varied across the hydrologic regime. Grasses in the xeric sites used a higher percentage of groundwater than grasses in mesic sites. This pattern was correlated with aboveground productivity, with grasses in the drier sites producing more aboveground biomass compared to mesic site grass production.
