Date of Award
Open Access Dissertation
High Frequency (HF) radar systems are commonly used to estimate surface ocean currents over the coastal ocean. Their range depends on their operational frequency and low frequency systems (≤ 10 MHz) can reach distances up to 200 km from the coastline. These systems are used to estimate surface currents by measuring the phase speed of wind-driven waves and comparing the measured speed with that expected theoretically; deviations from the theoretical still-water phase speed are attributed to ocean surface currents. Although HF radar systems are considered a mature technology and the accuracy of the radar-derived surface current estimates is well studied, the theoretical phase speed of wind-driven waves and its dependence on sea state is still unresolved. Additionally, the algorithm widely used for estimating surface currents for compact cross loop systems, the most commonly used HF radar systems, has not been adopted for use with beamforming linear array systems. Lastly, although different large scale HF radar networks have publicly available data (e.g., HFRnet in the USA hosts measurements from over 100 operational HF radar sites), there is a lack of openly available toolsets (such as eddy identification routines) to exploit these data sets for ocean research.
A 7-month data set from two HF radar sites [CSW and GTN, located on the coastline surrounding Long Bay, SC (USA)] is used to assess the performance of three different algorithms for estimating surface ocean currents from two linear array HF radar systems. The delay-and-sum beamforming algorithm, commonly used with beamforming systems, is compared with two direction finding algorithms, MUltiple Signal Classification (MUSIC) and direction finding using beamforming (Beamscan). While all three algorithms perform well along the radar boresight directions (𝑅2≈0.8), at ≈40∘ from their boresights, a baseline comparison between the two HF radar sites results in Beamforming performing poorly (𝑅2=0.01), MUSIC (𝑅2=0.37) performing better, while Beamscan (𝑅2=0.76) performed well.
Although use of the Beamscan algorithm can increase the accuracy of HF radar measured ocean wave phase speeds at high angles from the radar boresight directions, conversion of this information to sea surface currents requires understanding the nature of this measurement. If the phase speed of the ocean waves the radar is measuring is modified by the sea state itself, we need to account for it. Recent efforts to quantify the phase speed dependency of wind driven waves in different sea states have led to the development of three approaches that describe the fraction of the Stokes drift contributing to wave phase speed modification: a nonlinear weighted and depth averaged Stokes drift (effective Stokes drift), the Stokes drift mostly from longer waves than the wave in question (filtered Stokes drift), and half the surface Stokes drift.
Using the same 7-month data set from the two HF radars in 2017, filtered Stokes drift shows the best correlation to the difference between the radar and in situ current measurements (accounting for >60% of the variance between the two) and it could be used to correct HF radar derived surface velocities. With that in mind, a neural network method is developed that uses the amplitude of the four Bragg peaks from these two radar systems to predict the filtered Stokes drift terms. The filtered Stokes drift prediction correlates to the difference between the radar and in situ measurements with 𝑅2>0.6, reducing the difference between the in situ measurements and HF radar measurements significantly.
Finally, an optimized eddy identification routine (the winding-angle method) is presented for use with HF radar surface current data. Using data from the same two radar sites collected over 2013 the new method was utilized. It successfully identified over 1000 eddies with more than 200 of them being able to be tracked for periods in excess of 6 hours. All but 1 of the eddies tracked for over 48 hours were cyclonic upwelling eddies on the shelf break, representing the Charleston Gyre. We detail the passage of an anticyclonic eddy near the shelf moving equatorward and its associated momentum flux across the shelf break. We also find that the shelf break eddies contribute to momentum flux across the shelf and are most commonly spun up during times when wind is from the north, which induces Ekman transport in the preferred propagation direction of these anticyclonic shelf break eddies.
All codes developed as part of this dissertation described in the relevant chapters are publicly available at zenodo.org.
Cahl, D.(2023). HF Radar: Shining a Light on Ocean Currents. (Doctoral dissertation). Retrieved from https://scholarcommons.sc.edu/etd/7434