We study the adaptive sensing problem for the multiple source seeking problem, where a mobile robot must identify the strongest emitters in an environment with background emissions. Background signals may be highly heterogeneous, and can mislead algorithms which are based on receding horizon control, greedy heuristics, or smooth background priors. We propose AdaSearch, a general algorithm for adaptive sensing. AdaSearch combines global trajectory planning with principled confidence intervals in order to concentrate measurements in promising regions while still guaranteeing sufficient coverage of the entire area. Theoretical analysis shows that AdaSearch significantly outperforms a uniform sampling strategy when the distribution of background signals is highly variable. Simulation studies demonstrate that when applied to the problem of radioactive source-seeking, AdaSearch outperforms both uniform sampling and a receding time horizon information-maximization approach based on the current literature. We corroborate these findings with a hardware demonstration, using a small quadrotor helicopter in a motion-capture arena.