| 27 Jun 2025
Temporal and spatial scales of Near-Intertial Oscillations inferred from surface drifters
Abstract. The study investigates near-inertial oscillations (NIOs) in ocean surface currents, which are important for understanding air-sea interactions and improving satellite measurements of ocean currents, such as those planned by the ODYSEA mission. Traditional methods for measuring near-surface currents, like drifters and HF radars, have limited spatial coverage, while satellite altimetry only captures geostrophic currents, missing high-frequency components like NIOs.
Using a decade-long hourly drifter dataset (2010–2021) and outputs from the high-resolution coupled Ocean/Atmosphere model LLC2160, the study estimates the global spatial decorrelation length scales of NIOs. Drifter pairs were analyzed by latitude and distance, isolating the inertial frequency band. The LLC2160 model, which includes tidal and wind-driven forcing, was validated against the drifter data, showing good agreement in spectral characteristics and correlation scales.
Results show that NIO signals have larger decorrelation scales (~105–110 km) compared to low-frequency currents (~70–75 km), with variations by latitude (down to 50 km at high latitudes and near 20° N). These findings suggest that NIO energy is spread over larger spatial scales, linked to atmospheric forcing patterns. The LLC2160 simulation reliably reproduces these spatial characteristics and thus serves as a realistic tool for supporting ODYSEA mission planning.
The study shows that understanding NIO spatial coherency is critical for interpreting satellite Doppler measurements and mitigating aliasing effects, thereby improving ocean current observation capabilities.
My name is Shane Elipot and I am happy to provide a review of “Temporal and spatial scales of Near-Intertial Oscillations inferred from surface drifters” by Etienne et al.
This study could have given us some valuable insights into the spatial scales of near inertial oscillations (NIOs) in the ocean. But, unfortunately, the results aren’t very strong and don’t offer much in terms of interpretation. I think a major revision is needed before it can be published.
First, I noticed the title might be a bit misleading. Could you please clarify where the temporal scales of NIOs are presented in the paper? It seems like only estimates of spatial scales are shown in Figure 5. Apart from the intrinsic time scales of NIOs determined by the inertial frequency, I haven’t found any results about their decay temporal time scales or any other time scales. By analyzing single trajectories, this is a topic I explored in my own study from 2010 (Elipot et al. 2010 doi: 10.1029/2009JC005679). In this study, I also demonstrate the significant influence of the mesoscale on the characteristics of NIOs, which you seem to dismiss as “certainly not dominant”. Could you please provide the basis for this statement?
Second, I’m curious about the method you’re using to interpolate model velocities onto the position and time (of year?) of the drifters. Is this really a way to validate the model (LLC2160-C1440) for simulating NIOs? Because the displacements of the true drifters don’t match the velocity field of the model, their pair separation distances don’t either. In fact, you’re sampling the model in a way that’s pretty limited. Alternatively, you could use all the data points from the model to calculate the spatial decorrelation scale of NIOs from filtered model velocities. That way, I think the studies by Yu et al. 2019 and Arbic et al. 2022 (doi: 10.1029/2022JC018551) are a more straightforward comparison of NIOs in drifter data and in model data. Even though these studies are only concerned with the energy of NIOs, not their temporal or spatial scales, which is what you’re interested in.
Third, the main results in Figure 5 might be reliable, but they’re not presented in a robust way. It’s important to include uncertainty estimates for these results. How can we be sure that the spatial scales are different between NIO and low-frequency motions without any uncertainty estimates? Also, the paper should mention how many pairs of data are actually used in the calculations. Since the size of your boxes and the distribution of the data change depending on latitude, I think the uncertainty estimates in Figure 5b would also vary a lot. Additionally, I think some of the choices of the analyses aren’t explained well. Why would a correlation threshold of 0.5 give us a meaningful decorrelation scale estimate? How dependent are the results in Figure 5b to this choice? Other metrics could be used, like the x-axis intercept of the slope at distance zero, or fitting parametric models of the decorrelation or cross-covariance function to your results.
Another technical comment is about the filtering method. My understanding is that you filter the velocity in the range [0.9 f, 1.1 f] where f is the inertial frequency at the center latitude of each 5 degree latitudinal band. (As I understand what you mean by “f computed for the mean latitude of the 5° bin”, or do you mean the mean value of the drifter positions in that band?). If my understanding is correct, an simple calculation shows that [0.9 f, 1.1 f] covers the range of inertial frequencies in a 5 degree band only for latitudes above 22.5 degrees. Below that latitude, the range does not cover the full range of inertial frequencies and you might be underestimating the inertial velocities. Also, why presenting only the zonal velocity autocorrelation functions? If you considered the phase and amplitude of the complex-valued velocity autocovariance function you would disentangle the decay of the covariance due to phase on one hand, and due to amplitude on the other hand.
Finally, I noticed that there hasn’t been much discussion about the meaning and importance of your results. I’m curious to know how these results can be explicitly applied to the potential data from the upcoming ODYSEA satellite mission.
I attach a PDF document with some detailed comments, probably best viewed with Adobe Acrobat.