Improving tidal accuracy in a high-resolution global ocean circulation model
Improving tidal accuracy in a high-resolution global ocean circulation model
Abstract:
The ocean tidal velocity and elevation can be estimated concurrently with the ocean circulation by adding the astronomical tidal forcing, parameterized topographic internal wave drag, and self-attraction and loading to the general circulation physics. We have previously developed an augmented state ensemble Kalman Filter (ASEnKF) approach that improves the accuracy of the barotropic tides embedded in a 1/12.5° three-dimensional ocean general circulation model. In the ASEnKF, an extra term, meant to correct for errors in the tide model due to imperfectly known topography and damping terms, is introduced into the barotropic tidal forcing. The ASEnKF method yields an optimal estimate of the model forcing correction terms, which can be applied to the fully three-dimensional ocean model. The deep-water, global area-averaged RMS sea surface elevation error of the principal lunar semidiurnal tide M2 was reduced from 4.4 cm in a best-case non-assimilative solution to 2.6 cm. The largest elevation errors in both the nonassimilative and ASEnKF solutions are in the North Atlantic, a highly resonant basin. The perturbations used in our first ASEnKF effort were large scale in nature and randomly distributed. The large-scale perturbations did not have significant responses in these resonant regions because they did not represent the smaller scale errors we were trying to reduce. We have an effort underway that builds on this earlier work, using perturbations that vary in their size and location, targeting resonant areas with large tidal errors such as Hudson Strait, English Channel, Patagonia and the Amazon Delta. Experiments illustrate both local and remote impacts of these limited area perturbations and exhibit variance that is collocated with areas of large tidal error, a necessary ingredient to obtain the desired corrections. Work is underway to complete the running of a perturbed ensemble set from which a new forcing correction can be derived. We’ll present an update on this effort and our plans for future development.