Generalized inversion of the Gent-Cane model of the tropical Pacific with Tropical Atmosphere-Ocean (TAO) data
Abstract:We here describe the results of our latest effort to reanalyze TAO monthly mean surface and subsurface temperature observations constrained by a tropical Pacific ocean model, and simultaneously to evaluate the physical consistency of the observations and the model. Both tasks are executed by weak–constraint, four–dimensional variational assimilation of the observations into the model. In this study our reanalysis employs the reduced–gravity Gent–Cane model, combined with the 'KPP–interior' parameterization of vertical turbulent fluxes. With the limited vertical resolution adopted in previous studies of this model, the 'W4DVAR' or inverse method fails to produce an acceptable reanalysis, as the dynamical residuals are too large locally in space and time. Moreover, the objective significance test that is an essential product of the inversion rejects the model, even though the model is imposed only as a weak constraint, as convincingly as the even simpler Zebiak–Cane model was rejected (Bennett et al., 1998, 2000). In order to obtain a locally plausible reanalysis of the observations, we have to employ significantly finer vertical resolution. The calculations are extremely expensive and technically difficult; in the interests of efficiency and convergence, we solve the fixed–interval smoothing problem for 3–month intervals (during the 1997–1998 El Niño, December 1996–March 1998), and we impose the continuity equations for layer thickness as strong constraints. We present detailed results for just one such experiment which shows the model in the best light. The resulting fits to the monthly-mean data are within our prior error assumptions and so constitute a highly plausible reanalysis. However, the significance test statistic for the inversion exceeds its expected value by many tens of standard deviations, forcing us to the inference that the model cannot in fact be reconciled with the observations. The paradox of locally small residuals and a large test statistic is explained by an analysis of the degrees of freedom allowed in the weak constraints.
Document Type: Research Article
Publication date: January 1, 2006
More about this publication?
- The Journal of Marine Research publishes peer-reviewed research articles covering a broad array of topics in physical, biological and chemical oceanography. Articles that deal with processes, as well as those that report significant observations, are welcome. In the area of biology, studies involving coupling between ecological and physical processes are preferred over those that report systematics. Authors benefit from thorough reviews of their manuscripts, where an attempt is made to maximize clarity. The time between submission and publication is kept to a minimum; there is no page charge.
- Editorial Board
- Information for Authors
- Subscribe to this Title
- ingentaconnect is not responsible for the content or availability of external websites