Nonlinear noise reduction for the observation data of climatology based on the searching average method

Abstract

The searching average nonlinear noise reduction method, which is based on local linear fit to the nonlinear dynamics, is introduced to reduce the noise in the observation data of climatology. Recurrence plots are used to estimate the size of local neighbors. The noise reduction is improved markedly. In order to show the validity of the program in noise reduction, it is first applied to a noise time series of Henon map contaminated by Gaussian white noise. And then, this noise reduction scheme is applied separately to the observation data of meteorology. The analyses of a nonlinear prediction demonstrate the efficiency of the method for noise reduction.