Sensing uncertainty is a key performance metric of interest to any application based on a sensor network. It is thus desirable to sense the phenomenon of interest at the highest possible sensing resolution to achieve the least uncertainty. Given a finite set of sensing resources, it is of interest to determine how to allocate these resources for the best sensing performance. This allocation depends on the distribution of the sensed phenomenon in space and also the characteristics of the medium, such as the presence of obstacles. We propose methods to acquire this environmental information and then move the sensors to improve the sensing performance.