[Paper Review] High-Frequency Radar Ocean Current Mapping at Rapid Scale with Autoregressive Modeling

Baptiste Domps, Dylan Dumas+2 more|arXiv (Cornell University)|Jun 19, 2020

Oceanographic and Atmospheric ProcessesEarth and Planetary Sciences35 references21 citations

TL;DR

This paper proposes an autoregressive modeling approach combined with the Maximum Entropy Method (AR-MEM) to enhance high-frequency radar (HFR) ocean current mapping at rapid temporal scales. By replacing conventional FFT-based spectral estimation with AR-MEM, the method achieves reliable radial current estimates in just one minute of integration time—dramatically improving coverage and signal-to-noise robustness—while revealing turbulent spectral decay consistent with Kolmogorov's −5/3 power law at minute-scale resolution.

ABSTRACT

We use an Autoregressive (AR) approach combined with a Maximum Entropy Method (MEM) to estimate radial surface currents from coastal High-Frequency Radar (HFR) complex voltage time series. The performances of this combined AR-MEM model are investigated with synthetic HFR data and compared with the classical Doppler spectrum approach. It is shown that AR-MEM drastically improves the quality and the rate of success of the surface current estimation for short integration time. To confirm these numerical results, the same analysis is conducted with an experimental data set acquired with a 16.3 MHz HFR in Toulon. It is found that the AR-MEM technique is able to provide high-quality and high-coverage maps of surface currents even with very short integration time (about 1 minute) where the classical spectral approach can only fulfill the quality tests on a sparse coverage. Further useful application of the technique is found in the tracking of surface current at high-temporal resolution. Rapid variations of the surface current at the time scale of the minute are unveiled and shown consistent with a $f^{-5/3}$ decay of turbulent spectra.

Motivation & Objective

• To address the limitation of conventional HFR signal processing, which relies on long integration times (10–60 min) for reliable current estimation.
• To improve the reliability and spatial coverage of radial surface current estimation under short integration times.
• To enable high-temporal-resolution monitoring of rapid surface current variations for emerging applications such as tsunami early warning and oil spill tracking.
• To validate the AR-MEM method against both synthetic data and real-world HFR measurements in Toulon, France.
• To investigate the presence of turbulent spectral behavior at minute-scale temporal frequencies using high-resolution current estimates.

Proposed method

• The method models the complex voltage time series from HFR radar as an autoregressive (AR) process to capture underlying spectral dynamics.
• The AR coefficients are estimated using the Maximum Entropy Method (MEM), which provides higher spectral resolution and better noise suppression than standard FFT.
• The AR-MEM approach enhances the amplitude of the first-order Bragg peaks in the Doppler spectrum, improving detection of radial current velocity.
• The technique is applied to synthetic HFR data to evaluate performance under varying signal-to-noise ratios and integration times.
• The method is validated on real experimental data from a 16.3 MHz WERA HFR system in Toulon, comparing AR-MEM results with classical FFT-based estimation.
• Power spectral density (PSD) analysis of radial current fluctuations is performed to detect turbulent scaling behavior, particularly the −5/3 power law.

Figure 1: Simulated bistatic ( $\displaystyle\varphi=15\degree$ ) second-order Doppler spectrum at radar frequency 16.3 MHz, for a Pierson-Moskowitz oceanic spectrum with wind speed $\displaystyle 5$ m.s -1 in the absence of surface current and noise. The first-order Bragg peaks are located at $\dis

Experimental results

Research questions

• RQ1 Can AR-MEM improve the accuracy and reliability of radial surface current estimation in short integration times (e.g., ~1 minute) compared to classical FFT-based spectral analysis?
• RQ2 Does the AR-MEM method significantly increase the spatial coverage of valid current estimates under low signal-to-noise conditions?
• RQ3 Can the AR-MEM technique resolve rapid temporal variations in surface currents at the minute timescale, enabling high-temporal-resolution monitoring?
• RQ4 Is the power spectral density of radial current fluctuations consistent with a −5/3 decay, indicative of Kolmogorov turbulence, at high temporal frequencies?
• RQ5 Can AR-MEM-based current maps be validated against independent in-situ drifters' data?

Key findings

• The AR-MEM method achieves reliable radial current estimation with integration times as short as 1 minute, where the classical FFT method fails to meet quality criteria over more than 90% of the radar coverage.
• For short integration times (e.g., 1 min), the AR-MEM method increases the success rate of current estimation by over 50% compared to FFT, particularly in low signal-to-noise conditions.
• The AR-MEM method enables the detection of turbulent spectral behavior with a −5/3 power law decay in the radial current PSD, extending over at least two decades of temporal frequency.
• The power spectral density (PSD) of radial currents estimated with AR-MEM extends to higher frequencies (up to ~6.7 cycles per hour) than FFT, which saturates at the Nyquist limit.
• Experimental validation with in-situ drifters confirms the high accuracy of AR-MEM-based current estimates, particularly at high temporal resolution.
• The AR-MEM approach reveals rapid current fluctuations at the minute timescale, demonstrating its potential for real-time applications such as tsunami early warning and Lagrangian transport modeling.

Figure 2: Influence of the AR order in estimating the radial current $\displaystyle U_{r}$ from a synthetic HFR time series with AR-MEM for an integration time of 4.5 min ( $\displaystyle N=1024$ time steps) and $\displaystyle U_{r}=0.3$ m.s -1 . The radar signal is modeled by a random stationary pr