We consider the problem of a sensor network tracking a moving target that exhibits a Markov model of mobility. The sensor nodes have adjustable power levels and the precision of the measurement of the target location depends on both the relative distance from the target to the measuring sensor, and on the sensing power level used by that sensor. An important issue in sensor networks is the power effi- ciency, thus we consider the optimization of a family of cost functions that include both the accuracy of a measurement and the power used to do that measurement. We define our problem as a control policy optimization for a partially observed Markov chain. For such scenarios, we derive optimal power control policies based only on partial observations of the target location, and propose hand-off techniques based on this policies.