Regional adjustment of inertial gravity disturbance vector measurements by optimal two-dimensional smoothing

Abstract

A two-dimensional signal processing algorithm is developed to obtain smoothed estimates of the gravity disturbance vector from vector measurements obtained by an inertial surveying system. The method differs from a conventional least squares regional adjustment of such measurements in that it accommodates a signal model in the smoothing process. Using principles from the physical theory of geodesy, it is shown that for a local region on the surface of the earth, an appropriate signal model is obtained by applying the two-dimensional Laplacian operator to a function representing the surface disturbance potential and equating the result to spatial white noise. The model of the vector measurement is the three orthogonal spatial derivatives of a three dimensional disturbance potential evaluated at the surface contaminated by additive white noise. The problem of simultaneous smoothing of all the gravity disturbance measurements from all survey traverses in the region is solved by representing the surface disturbance potential by a two-dimensional Karhunen-Loeve expansion that makes no specific reference to either the geometry or the ordering of the parameter space, thereby making no assumptions of causality, stationarity or isotropy. The problem of estimating the gravity anomaly and the two vertical deflection components reduces to estimating the Karhunen-Loeve coefficients which are uncorrelated and rapidly converging. Simulation results as well as smoothing of actual gravity disturbance vector measurements obtained by the U.S. Army Engineer Topographic Laboratories (USAETL) with the Rapid Geodetic Survey System (RGSS) at the White Sands Missile Range (WSMR) are presented in the paper. An analysis of these results shows that the optimal two-dimensional smoother obtains a performance benefit relative to conventional regional least squares by a factor of 2 and a benefit relative to single-traverse smoothed results by a factor of 4.