Application of particle swarm optimization method to incoherent scatter radar measurement of ionosphere parameters

Abstract

Simultaneous derivation of multiple ionospheric parameters from the incoherent scatter power spectra in the F 1 region is difficult because the spectra have only subtle differences for different combinations of parameters. In this study, we apply a particle swarm optimizer (PSO) to incoherent scatter power spectrum fitting and compare it to the commonly used least squares fitting (LSF) technique. The PSO method is found to outperform the LSF method in practically all scenarios using simulated data. The PSO method offers the advantages of not being sensitive to initial assumptions and allowing physical constraints to be easily built into the model. When simultaneously fitting for molecular ion fraction (f m ), ion temperature (T i ), and ratio of ion to electron temperature (γ T ), γ T is largely stable. The uncertainty between f m and T i can be described as a quadratic relationship. The significance of this result is that T i can be retroactively corrected for data archived many years ago where the assumption of f m may not be accurate, and the original power spectra are unavailable. In our discussion, we emphasize the fitting for f m , which is a difficult parameter to obtain. PSO method is often successful in obtaining f m , whereas LSF fails. We apply both PSO and LSF to actual observations made by the Arecibo incoherent scatter radar. The results show that PSO method is a viable method to simultaneously determine ion and electron temperatures and molecular ion fraction when the last is greater than 0.3.

Determination of the Signal Fluctuation Threshold of the Temperature‐Ion Composition Ambiguity Problem Using Monte Carlo Simulations

“…To avoid additional ACF conversion times at each fitting comparison, estimation analysis was done by directly comparing the difference between simulated and theoretical spectrums at each range (Wu et al, ). Theoretical ion acoustic spectra were obtained at frequencies between ±10 kHz.…”

“…In Figure the TICA effect is shown for two different spectrums obtained with two different combinations of parameters. Under certain conditions, this mixture of ion species might be determined unambiguously (Aponte et al, ; Lathuillere et al, ; Oliver, ; Wu et al, ). The determination of the conditions required for an unambiguous estimation is the main focus of this study.…”

“…The PSO algorithm obtained much better estimations than the standard NLLS algorithm commonly used in ISRs. Results from Lathuillere et al () and Wu et al () demonstrate the feasibility of unambiguously estimating plasma parameters of signals with very small fluctuations. Nevertheless, no previous study has assessed the required signal fluctuation level to unambiguously estimate the ion composition and temperature parameters in the TICA problem.…”

“…Recently, Wu et al () used a new optimization algorithm to unambiguously estimate plasma parameters of ISR signals with very small fluctuations. In their work, signals with high signal‐to‐noise ratio (SNR) values, in the range of 15 to 25, were obtained at the Arecibo Radio Observatory using simultaneous frequency transmission (Sulzer, ) and coded long pulse (Sulzer, ) techniques.…”

“…In their work, signals with high signal‐to‐noise ratio (SNR) values, in the range of 15 to 25, were obtained at the Arecibo Radio Observatory using simultaneous frequency transmission (Sulzer, ) and coded long pulse (Sulzer, ) techniques. Wu et al () used the particle swarm optimization (PSO) algorithm (Kennedy & Eberhart, ) to analyze ISS data with the addition of a priori information from the plasma line. The PSO algorithm obtained much better estimations than the standard NLLS algorithm commonly used in ISRs.…”