Dr Daniel B. Whitt, Nicholson S., Dr Magdalena Carranza

Subseasonal surface wind variability strongly impacts the annual mean and subseasonal turbulent atmospheric surface fluxes. However, the impacts of subseasonal wind variability on the ocean are not fully understood. Here, we quantify the sensitivity of the ocean surface stress (τ), buoyancy flux (B) and mixed‐layer depth (MLD) to subseasonal wind variability in both a one‐dimensional (1D) vertical column model and a three‐dimensional (3D) global mesoscale‐resolving ocean/sea‐ice model. The winds are smoothed by time‐filtering the pseudo‐stresses, so the mean stress is approximately unchanged and some important surface flux feedbacks are retained. The 1D results quantify the sensitivities to wind variability at different timescales from 120 days to 1 day at a few sites. The 3D results quantify the sensitivities to wind variability shorter than 60 days at all locations, and comparisons between 1D and 3D results highlight the importance of 3D ocean dynamics. Globally, subseasonal winds explain virtually all of subseasonal τ variance, about half of subseasonal B variance, but only a quarter of subseasonal MLD variance. Subseasonal winds also explain about a fifth of the annual mean MLD and a similar and spatially‐correlated fraction of the mean friction velocity, urn:x-wiley:21699275:media:jgrc23678:jgrc23678-math-0001 where ρsw is the density of seawater. Hence, the subseasonal MLD variance is relatively insensitive to subseasonal winds despite their strong impact on local B and τ variability, but the mean MLD is relatively sensitive to subseasonal winds to the extent that they modify the mean u*, and both of these sensitivities are modified by 3D ocean dynamics.

Link to Full Article

Annual mean (top) and seasonal cycle amplitude (bottom) of the magnitude of the ocean mixed-layer depth from the control run (CTL, left) and the fractional difference between CTL and a low-pass run with smoothed winds (CTL-LP, right; see section 2.5 for more on the metrics). In (B) and (D), points are blanked if zero is included in the 95% confidence interval, which is derived non-parametrically using 1000 bootstrap samples at each point. In (B) and (D), red means the metric is greater in CTL, whereas blue means the metric is greater in LP.