|Title||Toward realistic nonstationarity of semidiurnal baroclinic tides in a hydrodynamic model|
|Publication Type||Journal Article|
|Year of Publication||2019|
|Authors||Nelson A.D, Arbic B.K, Zaron E.D, Savage A.C, Richman J.G, Buijsman M.C, Shriver J.F|
|Type of Article||Article; Early Access|
|Keywords||altimetry; comparison; data set; general-circulation; HYCOM; Internal tide; internal-tide; mesoscale; model-data; nonstationary tide; ocean; oceanography; reflection; surface manifestation; SWOT; variability|
Semidiurnal baroclinic tide sea surface height (SSH) variance and semidiurnal nonstationary variance fraction (SNVF) are compared between a hydrodynamic model and altimetry for the low- to middle-latitude global ocean. Tidal frequencies are aliased by similar to 10-day altimeter sampling, which makes it impossible to unambiguously identify nonstationary tidal signals from the observations. In order to better understand altimeter sampling artifacts, the model was analyzed using its native hourly outputs and by subsampling it in the same manner as altimeters. Different estimates of the semidiurnal nonstationary and total SSH variance are obtained with the model depending on whether they are identified in the frequency domain or wave number domain and depending on the temporal sampling of the model output. Five sources of ambiguity in the interpretation of the altimetry are identified and briefly discussed. When the model and altimetry are analyzed in the same manner, they display qualitatively similar spatial patterns of semidiurnal baroclinic tides. The SNVF typically correlates above 80% at all latitudes between the different analysis methods and above 60% between the model and altimetry. The choice of analysis methodology was found to have a profound effect on estimates of the semidiurnal baroclinic SSH variance with the wave number domain methodology underestimating the semidiurnal nonstationary and total SSH variances by 68% and 66%, respectively. These results produce a SNVF estimate from altimetry that is biased low by a factor of 0.92. This bias is primarily a consequence of the ambiguity in the separation of tidal and mesoscale signals in the wave number domain.