An Improved Iterative Algorithm Utilized in Data Processing for Incoherent Scatter Radar

Abstract

Measurements by incoherent scatter radars (ISRs) play very important roles in modern ionospheric and geospace studies. The ionospheric parameters are normally extracted from echo signals of ISR using nonlinear iteration algorithms. The processing speed and accuracy significantly rely on the iteration algorithm. The Levenberg-Marquardt (LM) method is a widely used algorithm for nonlinear iteration in modern ISR data processing. In the present letter, a new iteration algorithm, which is the so-called dog leg (DL) method, was introduced. Comparing with the LM method, the step size for the iteration with the DL method is not fixed. The iteration step size for the DL method keeps being adjusted in accordance with the trust region during the iteration process. The performance of the DL method was compared with that of the LM method using the simulation tests of ionospheric parameter extraction from ISR power spectra. The results showed that the DL method has better accuracy and iteration speed than that of the LM method. In addition, the error of the DL method is more stable than that of the LM method. Moreover, the DL method is more convergent than the LM method, which helps avoid the lack of ionospheric parameter due to the divergence of iteration with the LM method. Therefore, the DL method could be a more appropriate iteration algorithm for data processing of ISR measurements.