An efficient algorithm for the large-scale smoothing of scattered data retrieved from remote sounding experiments

Abstract

We present a new algorithm for smoothing/interpolation of two-dimensional fields applicable to noisy data observed at scattered sites. The technique is based on a special statistics allowing one to simultaneously minimize the fit residual and the correlation between residuals of adjacent points. The principle of the method is first explained in the 1-D case and then extended to the 2-D case by adjunction of a regularization operator. The method is compared with different algorithms (Loess-Renka, Biharmonic Spline and kriging) in three test cases related to remote sounding of the Earth's atmosphere by space-borne experiments.