In this paper, optimal sensing strategies are presented which use spatial averaging of noisy measurements of an advection-diffusion process for parameter estimation to trace the maximum of a field of interest. Each strategy utilizes nonlinear least-square (NLS) optimization to estimate the parameters of a locally modeled advection-diffusion process. In the first strategy, nonholonomic sensing agents are steered in the direction that maximizes an objective function. Using ideas from the first strategy, a second strategy is designed to find the maximum of a complicated advection-diffusion process in a non-convex surveillance region. Each time the NLS optimization is run, the target position for the nonholonomic sensing agents is updated with the estimated maximum concentration position of the locally modeled advection-diffusion process. To ensure the nonholonomic sensing agents do not collide with boundaries in the non-convex region, an object avoidance algorithm is utilized.