The ``large-scale'' is used here to refer to scales larger than 1000 km, a loosely defined upper limit for the mesoscale discussed in Section 4. In this range of scales, the signal-to-noise ratio is generally small for all the altimeters except for TOPEX/POSEIDON. As noted above, a variety of error sources such as the orbit, the tides, the signal transmission media, and the instrument systematic errors have dominant scales larger than 1000 km. Despite these obstacles, numerous attempts have been made to examine the large-scale ocean variabilities using the pre-TOPEX/POSEIDON altimeter data. The dominant orbit errors were reduced by a variety of methods, ranging from a dynamic approach utilizing gravity model adjustment [ Koblinsky et al., 1992] to a purely geometric approach utilizing the spectral characteristics of the orbit error [ Wunsch, 1991 a]. The reader is referred to Wagner and Tai [1994] for a discussion of the effects of the orbit error removal procedures on the detection of ocean signals.
Comparison of the low degree and order spherical harmonics of the Geosat sea level variations to a set of tide gauge data yielded a rms difference of 5--10 cm [ Wunsch, 1991 a; Koblinsky et al., 1992]. Similar results were obtained by Harangozo et al. [1993]. Wunsch [1991 a] demonstrated a method of combining altimeter data with tide gauge data in enhancing the accuracy of the estimate of the global sea level variability. The results were used to investigate the large-scale response of the ocean to atmospheric forcing [ Wunsch, 1991 b]. Up to 50% of the variance was accounted for by the forcing of wind and atmospheric pressure. Spatial pattern of the global annual and semiannual variabilities were examined in detail by Koblinsky et al. [1992] and Jacobs et al. [1992]. Contamination of the annual cycle by ocean tides was removed by an empirical method in the latter study. The interannual change due to the 1986--87 El Niño were demonstrated [ Wunsch, 1991a; Koblinsky et al., 1992; Koblinsky, 1993]. Preliminary results from TOPEX/POSEIDON have demonstrated its capability of more accurate detection of the large-scale variabilities such as a hemispheric asymmetry in the global annual cycle, with the amplitude in the Northern Hemisphere twice as large as in the Southern Hemisphere [ Stammer and Wunsch, 1994; Cheney et al., 1994; Knudsen, 1994]. Behringer [1994] reported close agreement of the large-scale variabilities of the North Atlantic from the TOPEX/POSEIDON observations with an ocean data assimilation system within 2 cm.
There were also regional studies of the gyre- and basin-scale variabilities. The annual and interannual variabilities of the Alaska Gyre was investigated by Bhaskaran et al. [1993] based on a four-year Geosat database. He found significant correlations of the sea level with atmospheric pressure and the Southern Oscillation Index (an indicator for the climatic condition of the tropical Pacific based on the sea level pressure difference between Darwin and Tahiti). Kelly et al. [1993] examined the interannual variations of the Alaska and the California currents and found that the two currents were out of phase with each other. Chelton et al. [1990] examined the large-scale structure of the temporal variability of the Antarctic Circumpolar Current (ACC) and reported that only a minor portion (33%) of the variance was zonally coherent, corresponding to the annual, semiannual, and interannual variabilities. The current has strong regional characteristics in each basin. Using the Seasat altimeter and scatterometer data, Mestas-Nunez et al. [1992] analyzed the relation between the variability of the ACC and the wind forcing and reported that a quarter of the variance was accounted for by the linear response of the ocean to the forcing of the curl of the wind stress (the time-dependent Sverdrup relation). Gille [1994] used the technique of Kelly and Gille [1990] for estimating the mean dynamic topography of the entire ACC. Her results were also able to account for 40--70% of the sea level variance along the axis of the ACC in terms of the meandering of the current.
By comparing the global sea level derived from Seasat to that from
Geosat, Haines et al. [1992] examined the difference between the
summer of 1978 to that of 1987 and found good agreement with
in-situ observations. This result demonstrated the potential of
combining observations from different altimetry missions to monitor
long-term changes in the ocean. The success of Haines et al. [1992]
was largely due to their recomputing the Seasat orbit using the same
gravity model as the one used for Geosat, thus eliminating the large
difference in the geographically-correlated orbit errors that could be
aliased to temporal changes [ Cheney and Douglas, 1988]. Based on
the difference between the Geosat and the ERS-1 observations, Jacobs
et al. [1994] reported a large-scale sea level change from 1987 to
1992 in the Kuroshio Extension region. Using a model of the Pacific
Ocean, they demonstrated that the observed sea level change was
caused by a Rossby wave excited by the 1982--83 El Niño. It
took ten years for the Rossby wave to travel across the Pacific at
about 40
N. This study provides the first evidence for a
decadal effect of El Niño.
Change in the global mean sea level has both scientific and practical
implications. The dense, global sampling of a satellite altimeter
makes it one of the most viable approaches to monitoring the change
in the global mean sea level. Born et al. [1986] first demonstrated the
idea using the Seasat data. Due to the large orbit error, the result
showed more than 10 cm fluctuations in 24 days. Tapley et al. [1992]
showed that the rms variability of the global mean sea level calculated
from two years' worth of the Geosat data was about 2 cm. The rate
of change was estimated 
0.5 cm/year. Similar results were
obtained by Visser et al. [1993]. The various errors affecting the
global mean sea level estimate were discussed in detail by Wagner and
Cheney [1992]. Based on the first year of the TOPEX/POSEIDON
data, Nerem et al. [1994] reported that the rms variability of the
global mean sea level was reduced to a level of 0.5 cm. Although
significant progress has been made, there is still a long way before
altimeter can detect the 1 mm per year global mean sea level change.