Spectral analysis of highly aliased sea‐level signals

Abstract

Observing high‐wavenumber ocean phenomena with a satellite altimeter generally calls for “along‐track” analyses of the data: measurements along a repeating satellite ground track are analyzed in a point‐by‐point fashion, as opposed to spatially averaging data over multiple tracks. The sea‐level aliasing problems encountered in such analyses can be especially challenging. For TOPEX/POSEIDON, all signals with frequency greater than 18 cycles per year (cpy), including both tidal and subdiurnal signals, are folded into the 0–18 cpy band. Because the tidal bands are wider than 18 cpy, residual tidal cusp energy, plus any subdiurnal energy, is capable of corrupting any low‐frequency signal of interest. The practical consequences of this are explored here by using real sea‐level measurements from conventional tide gauges, for which the true oceanographic spectrum is known and to which a simulated “satellite‐measured” spectrum, based on coarsely subsampled data, may be compared. At many locations the spectrum is sufficently red that interannual frequencies remain unaffected. Intra‐annual frequencies, however, must be interpreted with greater caution, and even interannual frequencies can be corrupted if the spectrum is flat. The results also suggest that whenever tides must be estimated directly from the altimetry, response methods of analysis are preferable to harmonic methods, even in nonlinear regimes; this will remain so for the foreseeable future. We concentrate on three example tide gauges: two coastal stations on the Malay Peninsula where the closely aliased K1 and Ssa tides are strong and at Canton Island where trapped equatorial waves are aliased.