The net benefit of agroforestry trees for small scale farmers in dryland agricultural systems is debatable because while they provide significant direct and indirect services, they also consume considerable amounts of scare water resources. In this study we monitor the stable isotopes of water to improve a water budget of a Sclerocarya birrea tree in a millet field in South Eastern Burkina Faso. Data obtained from air temperature and humidity, surface temperature, solar radiation, and soil moisture sensors attached to a wireless sensor network uniquely configured around the agroforestry tree provided the initial calculation of the local water balance. Isotopic ratios were determined from water extracted from stems and sub canopy soil, and from nearby ground water, precipitation, and surface water that was sampled weekly. A linear mixing model is used to predict when the tree switched between water sources. The results from the linear mixing model coupled with a tree water balance demonstrate the extreme seasonality of the annual cycle of water use by this deciduous species.