Recent studies have illuminated the process of hydraulic redistribution, defined as the translocation of soil moisture via plant root systems, but the long-term ecohydrologic significance of this process is poorly understood. We investigated hydraulic redistribution (HR) by Prosopis velutina Woot. (velvet mesquite) in an upland savanna ecosystem over a two-year period. Our goal was to quantify patterns of HR by mesquite roots and assess how this affects tree water use and productivity. We used the heat ratio method to monitor bi-directional sap flow, an analog of HR, in both lateral and tap roots. Additionally, we monitored soil water content and used the eddy covariance technique to quantify ecosystem carbon dioxide and water exchange. Mesquite roots redistributed large amounts of water throughout the year, even during periods of canopy dormancy. Dormant season precipitation (November–March) was often taken up by shallow lateral roots and transferred downward in the soil profile by deeper lateral and tap roots. Such a transfer was also apparent when the trees were active and moisture from summer rainfall was plant available in the upper soil layers. As the upper soil layers dried, sap flow moving toward the canopy in the lateral roots diminished and water use from deeper soils increased via the taproots. The relationship between root sap flow and above-canopy fluxes suggested that deeper “stored” water from HR allowed the trees to transpire more in the spring that followed a winter with significant downward redistribution. Patterns of lateral and tap root sap flow also implied that redistribution may extend the growing season of the trees after summer rains have ended and surface soils are dry, thus allowing the trees to photosynthesize through periods of seasonal drought. The large hydrologic magnitude and the ecological effects of HR we studied, along with mounting evidence of this process occurring in many other ecosystems, indicates that HR should be accounted for in many ecohydrologic modeling efforts.