a b s t r a c tThe existing altimetric record offers an unprecedented view of sea level (f) variability on a global scale for more than 2 decades. Optimal inference from the data involves appropriate partition of signal and noise, in terms of relevant scales, physical processes and forcing mechanisms. Such partition is achieved here through fitting a general circulation model to altimeter and other datasets to produce a ''best'' estimate of f variability directly forced by the atmosphere-the signal of primary interest here. In this context noise comes primarily from instrument errors and meso-scale eddies, as expected, but spatial smoothing effectively reduces this noise. A separate constraint is thus formulated to measure the fit to monthly, large-scale altimetric variability that unlike the daily, pointwise constraint shows a high signal-to-noise ratio.The estimate is explored to gain insight into dynamics, forcing, and other factors controlling f variability. Contributions from thermo-steric, halo-steric and bottom pressure terms are all important depending on region, but slopes of steric spectra (red) and bottom pressure spectra (white) are nearly invariant with latitude. Much f variability can be represented by a seasonal cycle and linear trend, plus a few EOFs that can be associated with known modes of climate variability and/or with topographic controls. Both wind and buoyancy forcing are important. The response is primarily basin-bound in nature, but uneven patterns of propagation across basin boundaries are clearly present, with the Pacific being able to affect large portions of the Indian and Atlantic basins, but the Atlantic affecting mostly the Arctic.