Electromagnetic bias estimates based on TOPEX, buoy, and wave model data

Abstract

[1] For quantitative studies of the ocean using altimetric measurements, the sea surface height measurements must be corrected for the electromagnetic (EM) bias effect. Projectprovided EM bias correction algorithms were derived from the altimeter data as a function of wind speed and wave height through variance minimization techniques [Gaspar et al., 1994]. In this paper we characterize the impact of those corrections on the altimeter data and compare it with an empirical algorithm based on tower observations of wave slopes and wave age (W. K. Melville et al., Wave slope and wave age effects in measurements of EM bias, submitted to Journal of Geophysical Research, 2002) and with theoretical predictions of EM bias [Srokosz, 1986]. We find a significant correlation between high-frequency ocean signals with the project-provided EM bias correction. This suggests that an EM bias algorithm cannot be estimated from altimeter data alone through variance minimization techniques. Since the EM bias depends on parameters that cannot be measured by an altimeter (e.g., wave slope), improved theoretical corrections appear to be preferable to altimetry-based empirical estimates. Using wave buoy and wave model (WAM) data, we find good agreement between existing theoretical correction and empirical corrections based on wave slope and wave age information over a significant range of parameters. Although further work is needed to extend our tests of algorithms to larger wave slopes, existing wave slope and wave age-based algorithms appear comparable in goodness to project-provided empirical algorithms and may be used for routine altimetry processing in combination with WAM output on a global basis.

Impact of high‐frequency waves on the ocean altimeter range bias

“…Numerous recent papers highlight the dominance of the EM bias by the RMS slope which in turn has been shown to correspond to the bias predicted under WNL theory and, to some extent, under long‐wave short‐wave interaction theory. A stated implication [ Gommenginger et al , 2003; Kumar et al , 2003; Melville et al , 2004] is that a long‐wave focus is likely to be sufficient for on‐orbit corrections. However, this study finds that the two‐scale WNL theory appears to be relevant with the corresponding implication that both the long‐wave field cross‐skewness bias predictor and its attenuation by shorter waves matter.…”

“…Still there remains a need to improve this sea level correction as its uncertainty, of O (1%) of significant wave height, now looms large as other terms in the error budget are reduced. Moreover, there is a need to understand when a systematic temporal or spatial error in this correction is occurring [ Kumar et al , 2003; Chelton et al , 2001]. Thus a critical remaining need in sea state bias research is to understand when and why the bias varies.…”

New models for satellite altimeter sea state bias correction developed using global wave model data

“…Most recently, Kumar et al [2003] explored such comparisons. They investigated the applicability of the tower‐based Melville et al [2004] parameterisation of the EM bias on wave slope and wave age information to altimeter data.…”

“…Models based on direct power measurements instead of wind speed have also been proposed [ Melville et al , 1991; Scharroo and Lillibridge , 2005]. Recent consensus suggests that while this approach offers an effective and self‐contained (no dependence on external data sources) altimeter correction, this two‐parameter SSB model may not entirely parameterize range bias variability attributed to regional complexities in the ocean wave climate [ Glazman et al , 1994; Kumar et al , 2003].…”

“…While models created from satellite data have focused on addressing the degree of wave field development by including a wave age or pseudo wave age parameter to improve the bias estimates [ Fu and Glazman , 1991; Glazman et al , 1996], analyses derived from field experiment data and different theoretical modeling approaches have rather pointed out the likely impacts due to the presence or absence of swell [ Minster et al , 1992; Glazman et al , 1996; Branger et al , 1993] and to wave dynamics via tilt modulation, hydrodynamic modulation, and non‐Gaussian long wave surface statistics (e.g., surface skewness). These latter factors are all associated with higher order moments of the surface height power spectral density (PSD) compared to SWH which is related to the order 0 moment, such as the surface significant slope or wave steepness [ Parsons and Miller , 1990; Branger et al , 1993] obtained from the second moment of the surface height PSD, the long wave RMS slope [ Rodriguez et al , 1992; Chapron et al , 2001; Gommenginger et al , 2003; Millet et al , 2003a, 2003b; Melville et al , 2004; Kumar et al , 2003; Vandemark et al , 2005], the acceleration variance [ Elfouhaily et al , 1999, 2000, 2001] related to the fourth moment, the cross‐skewness quantity between long wave elevation and slope [ Vandemark et al , 2005], or such as finally orbital velocity [ Elfouhaily et al , 2001; Chapron et al , 2001; Millet et al , 2003a] related to the first moment of the surface height PSD.…”