Abstract's details
A promising parametric spectral analysis method applied to sea level anomaly signals
CoAuthors
Event: 2017 Ocean Surface Topography Science Team Meeting
Session: Quantifying Errors and Uncertainties in Altimetry data
Presentation type: Type Oral
Contribution: PDF file
Abstract:
Spectral analysis of sea level anomalies (SLA) is widely used in the altimetry community to understand the geophysical content of the measured signal, to assess and compare the outputs of different missions. Spectral content of SLA is used to characterize ocean at different scales and to estimate the instrumental noise. Based on the SLA spectrum, one can estimate the spectral slope at medium to large scales (relied to the Surface Quasi-Geostrophic (SQG) ocean dynamics theory) and the measurement noise (observed as a noise plateau at smallest scales).
A previous contribution [1] has pointed out the weaknesses of spectral analysis based on Fourier transform, mainly due to: (1) the convolutive bias which results in a biased estimation of the slope, the bias being related to the kind of observation weighting temporal window used and (2) the high variance of estimation leading to averaging several spectral estimations and raising the question of stationarity.
To overcome these two drawbacks, a parametric spectral analysis method is proposed. This method is based on Auto-Regressive (AR) modeling [2,3] which is known to provide a spectral estimation with a lower variance than the as outperforming Fourier-based methods in terms of variance, in the case of short observation windows, without any need for choosing a weighting temporal window. Moreover, in order to better match the SLA frequency contents on a log scale to match the log scale interest of the SLA frequency contents , warping is introduced as a preprocessing prior to spectral analysis as it is done in speech coding [4].
Comparisons between the proposed parametric method (called ARWARP) and classical Fourier Fourier-based methods have been performed on both simulated SLA signals obtained from theoretical spectra and real signals from a high-resolution altimeter SARAL/AltiKa at 40 Hz rate (Orbit – Range – Mean Sea Surface). Results on simulated SLA signals highlight the performance of the ARWARP method, in terms of bias and variance on spectral estimation. ARWARP can be applied on short segments of SLA signals, providing a local information of the ocean characteristics, which can be of promising use by the wider Cal/Val and altimetry science community.
[1] C. Mailhes & al., “Review of Spectral Analysis Methods Applied to Sea Level Anomaly Signals”, in Proc. Ocean Surface Topography Science Team Meeting (OSTST), La Rochelle, France, Oct. 31 - Nov. 4, 2016,https://meetings.aviso.altimetry.fr/fileadmin/user_upload/tx_ausyclsseminar/files/ERR_01_2016-11-03-9h-TESA-CNES-OSTST-UncertaintiesSession_9h00.pdf
[2] P. Stoica and R. Moses, Spectral Analysis of Signals, Prentice Hall, 2005.
[3] Steven M. Kay, Modern Spectral Analysis, theory and Applications, Prentice Hall, Signal Processing Series, 1988.
A previous contribution [1] has pointed out the weaknesses of spectral analysis based on Fourier transform, mainly due to: (1) the convolutive bias which results in a biased estimation of the slope, the bias being related to the kind of observation weighting temporal window used and (2) the high variance of estimation leading to averaging several spectral estimations and raising the question of stationarity.
To overcome these two drawbacks, a parametric spectral analysis method is proposed. This method is based on Auto-Regressive (AR) modeling [2,3] which is known to provide a spectral estimation with a lower variance than the as outperforming Fourier-based methods in terms of variance, in the case of short observation windows, without any need for choosing a weighting temporal window. Moreover, in order to better match the SLA frequency contents on a log scale to match the log scale interest of the SLA frequency contents , warping is introduced as a preprocessing prior to spectral analysis as it is done in speech coding [4].
Comparisons between the proposed parametric method (called ARWARP) and classical Fourier Fourier-based methods have been performed on both simulated SLA signals obtained from theoretical spectra and real signals from a high-resolution altimeter SARAL/AltiKa at 40 Hz rate (Orbit – Range – Mean Sea Surface). Results on simulated SLA signals highlight the performance of the ARWARP method, in terms of bias and variance on spectral estimation. ARWARP can be applied on short segments of SLA signals, providing a local information of the ocean characteristics, which can be of promising use by the wider Cal/Val and altimetry science community.
[1] C. Mailhes & al., “Review of Spectral Analysis Methods Applied to Sea Level Anomaly Signals”, in Proc. Ocean Surface Topography Science Team Meeting (OSTST), La Rochelle, France, Oct. 31 - Nov. 4, 2016,https://meetings.aviso.altimetry.fr/fileadmin/user_upload/tx_ausyclsseminar/files/ERR_01_2016-11-03-9h-TESA-CNES-OSTST-UncertaintiesSession_9h00.pdf
[2] P. Stoica and R. Moses, Spectral Analysis of Signals, Prentice Hall, 2005.
[3] Steven M. Kay, Modern Spectral Analysis, theory and Applications, Prentice Hall, Signal Processing Series, 1988.