Abstract
Estimates of solar normal mode frequencies from helioseismic observations can be significantly improved by using Multitaper Spectral Analysis (MTSA) to estimate spectra from time series, followed by wavelet denoising of the log spectra. MTSA leads to a power spectrum estimate with reduced variance and better leakage properties than the conventional periodogram. Under the assumption of stationarity, the log multitaper spectrum has a statistical distribution that is approximately Gaussian, making wavelet denoising an asymptotically optimal method for reducing noise. This study finds that a single $m-\nu$ spectrum benefits greatly from this pipeline, and that wavelet denoising alone can improve $m$-averaged spectra. Using GONG time series, we demonstrate that the number of modes successfully fitted increases by 20% to 60%, with the optimal number of tapers found to be fewer than 10 to avoid excessive broadening of narrow modes.
Introduction
Helioseismology relies on the precise measurement of solar p-modes to infer the Sun's internal structure and dynamics. These oscillations are typically captured as velocity time series. The standard method for spectral estimation, the periodogram, suffers from high variance and spectral leakage, which can obscure weak modes and complicate the extraction of physical parameters. This paper presents a robust signal-processing pipeline combining Multitaper Spectral Analysis (MTSA) and wavelet denoising to overcome these limitations and provide a clearer window into the solar interior.
Methodology
The proposed pipeline consists of two primary stages:
- Multitaper Spectral Analysis (MTSA): Instead of using a single window (taper) to look at the data, MTSA applies several orthogonal tapers—specifically Slepian or sine tapers—to the same time series. Each taper provides an independent estimate of the spectrum. Averaging these estimates significantly reduces variance and minimizes spectral leakage.
As the number of tapers ($K$) increases, the combined energy distribution becomes flatter, capturing more information from the time series while maintaining low leakage.
- Wavelet Denoising: The authors apply wavelet shrinkage to the logarithm of the multitaper spectrum. By moving to the log domain, the multiplicative noise of the power spectrum becomes additive and approximately Gaussian. This statistical transformation allows for the application of wavelet denoising, which is asymptotically optimal for Gaussian data, effectively removing background noise without shifting frequency peaks.
Results
The pipeline was evaluated using Global Oscillation Network Group (GONG) velocity data for angular degrees $l = 30, 65,$ and $100$. Key findings include:
- Higher Yield: The number of "good fits" (modes successfully identified by the GONG algorithm) increased by 20% to 60% for frequencies below 3.5 mHz compared to the periodogram.
- Better Convergence: The peak-fitting algorithm required 24% fewer iterations to converge on the multitaper spectra, indicating a smoother search landscape.
- Frequency Integrity: Crucially, the denoising process did not introduce systematic frequency shifts. The estimated frequencies, amplitudes, and widths remained consistent with traditional methods, except for very narrow modes where multitapering introduced slight broadening.
Discussion and Limitations
There is a fundamental trade-off in the number of tapers used. While more tapers reduce noise, they also broaden spectral peaks. The authors found that for helioseismic time series of this length, the optimal number of tapers is fewer than 10; exceeding this can cause narrow modes at low frequencies to become un-fittable. Additionally, at very high frequencies near the acoustic cutoff, wavelet denoising struggled to distinguish between naturally blended signal ridges and noise, sometimes reducing the number of good fits in that specific regime.
Conclusion
The combination of Multitaper Spectral Analysis and wavelet denoising offers a powerful tool for solar physics. By transforming the statistical nature of helioseismic noise and utilizing multiple independent spectral estimates, this approach significantly enhances the ability to resolve solar oscillation modes, providing a more accurate basis for inferring the Sun's internal structure.
Summary updated