Abstract
Development of a methodology for the optimal placement of multivariate sensors as an aid in the design of geophysical field experiments is shown. The optimal placement methodology relies on spatial correlation estimates, interpolation error estimates as provided by a multivariate optimal interpolation scheme, and optimization techniques using nonlinear programming. Atmospheric fields and their associated statistics are simulated by analytic functions to demonstrate the capabilities of the methodology. These include the ability to design new networks, to add sensors optimally to existing networks, and to place restrictions on the region in which sensors can be located by introducing physical and economical constraints on the nonlinear programming problem. It is demonstrated that the mean and variance of the interpolation error for all fields is generally smaller for analyses whose input is derived from optimal sampling locations rather than from subjectively chosen locations.
