JOURNAL OF GEOPHYSICAL RESEARCH,VOL. 99, NO. C12,PAGES24,867-24,881, DECEMBER 15, 1994 Ocean tides in the Asian semienclosed seas from TOPEX/POSEIDON P. Mazzega and M. Berg6 Unit6 Mixte de Recherche 39 / Groupe de Recherche de G6od6sie Spatiale, Toulouse, France Abstract. The eight leading ocean tides (M2, S2, N2, K2, K1, O1, P1, and Q1) have been mapped in Asian semienclosed seas by inverting combined sets of tide gauge harmonic constants and a reduced setof TOPEX/POSEIDON satellite altimeter data.The tidal mapsare given on a 0.5øx0.5 ø grid with their formal errorestimates. Numericalexperiments conducted in the South China Seahave shown the inverse solutions to be quite insensitive to changes in the parameters of the a priori covariance functions for both the tidal signals and data residual errors.Once the variousparameters of the inversionscheme are fixed, the algorithm is applied to the Sea of Okhotsk, the Seaof Japan, the "East China"Sea(including the Bo Hai andYellow Seas) andthe Indonesian Seas. A set of tide gauge constants, not used in our solutions, is then used for comparison. Though theseaccuracy estimates may be biasedbecause of the uneven stations coverage, we conclude thatthe inversion of only 21 cycles of TOPEX/POSEIDON data leads to solutions with an accuracy comparable to the Schwiderski (1980b, c) semiempirical model and the CartwrightandRay (1990) modelderived from Geosat altimetry. 1. Introduction Accurate models of the main ocean tidal constituents are ever more required by several observational as well as theoreticaldevelopments of geophysics. Since about 10 years it has been recognized that with the aim of reaching the needed accuracy, suchmodelscould not rely on the knowledge of the tidal astronomical potential and governing hydrodynamical equations only: direct observations of the tidal fields, up to now mainly in the form of tidal heights data, should be usedto constrainthe models. The semiempirical model produced by Schwiderski in 1980 [Schwiderski, 1980a]is the bestexample of a linearized hydrodynamical modelwith internalparameters adjusted in orderto reacha better fit with a large set of coastal tide gaugedata [Schwiderski, 1980b, c]. Nowadays,satelliteradar altimeters providea huge amount of measurements of the instantaneous ellipsoidalheight of the sea surface,all over the world ocean and during periodsof several years. Among other signals and noises, the altimeter datum includes the total tide. As much skill has to be developed for extractingthe ocean tides, it was only quite recently that satellite altimetry has achieved results comparable to theoreticalmodeling.Thus Cartwright and Ray [1990] obtained from the sole analysis of Geosat altimeter data a global model of the diurnal and semidiurnal complex admittances with an accuracy perhaps of the same orderof that of the Schwiderski model for the leading waves. During the same decade, assimilation or inverse methods have provided a rigorous theoretical frame to constrain the hydrodynamical models with observations [Bennett and Mc Intosh, 1982; Zahel, 1991; Jourdin, 1992]. Copyright 1994 by the American Geophysical Union. Papcr number 94JC01756. 0148-0227/94/94JC-01756505.00 The preliminary sea surface variabilities deducedfrom the TOPEX/POSEIDON missionunequivocally attest the high quality of the data from the two altimeters, the unprecedented accuracy of the reference radialorbits, andtheprogresses made in the various geophysical modelsused for corrections of the raw data.For the purpose of tidal studies, one hasto analyze the longest possibletime series of measurements so that the results presented in this issue should be considered as preliminary. Nevertheless, the incentives for trying to obtain the tidal heights from TOPEX/POSEIDON are numerous as will appear below. Before moving to the global inversion of TOPEX/POSEIDON data, we have decided to make numerical experiments at a relatively low computational cost, in domains of limited extensions. The Asian semienclosed seas are particularly interesting for our purpose. From a dynamical point of view, the tidal patterns are complex with several amphidromic systems andstrong gradients of the tidal ranges due to local resonances (see Ye and Robinson [1983] for a discussion of the tidal dynamics in the South ChinaSea). As a consequence the tidal regime may be predominantly diurnal or even mixed, a peculiarity difficult to graspwith linearized dynamical models, the friction and advection terms being almost equally driven by both semidiurnal and diurnal tidal velocityfields. From a phenomenological point of view, tides in these seasbear strongloading effects,as can be seen,for example, in the numerous gravity loadingmeasurements taken in the vicinity of some of these seas [Endo, 1985;Sun, 1992]. We alsodispose of a ratherlarge set of harmonic constants as derived from tide gauge measurements [International Hydrographic Office (IHO), 1979], divided into those to be inverted togetherwith satellite altimetry and those kept independent for models comparisons, as explained in the next section.The third sectiongives an overview of the inversion scheme. The variousnumerical testsperformed in the South China Sea are presented in section4, and then final results obtained in all the basins are presented in section 5. The overall performances of our inverse solutions and of the Schwiderski[1980a] and Cartwright and Ray [1990] models 24,867