Oceanic mixing must be sampled at the length and time scales of the mixing processes to evaluate the rates and mechanisms of the mixing with minimum ambiguity. Sampling is complicated by the wide range of scales and extreme intermittency of planetary turbulence, and the interpretation of microstructure data is subject to wide deviations between current theories of stratified turbulent mixing. The statistical and hydrodynamic issues are connected, but may be treated separately. Estimation of the mean mixing rate &khgr;¿ in a ''layer'' requires only that the probability density function of the local mixing rate &khgr; be known from measurements, whatever the processes leading to the mixing may be. The mixing may be either uniform or intermittent within the layer space-time control volume, and the flow may be turbulent or nonturbulent. Hydrodynamic questions arise because oceanic microstructure data sets from most layers indicate undersampling: the data generally do not include regions of strong active turbulence suggested by the large intermittency factors of the measured probability distribution functions. However, the data sets do include fossil-turbulence remnants of such events which confirm the undersampling hypothesis.

Baker and Gibson <1987> find that viscous and diffusive dissipation rates &egr; and &khgr; in most data sets have distributions indistinguishable from lognormal, with large intermittency factors &sgr;2ln&egr;,&khgr; in the range 3-7, where &sgr; is the variance about the mean. In a fossil-turbulence-mixing model, Gibson (1980-1987) shows that the microstructure measured is generally in a mixed fossil-and-active turbulence state, with previous active turbulence &egr; and &khgr; values large enough to be consistent with the measured intermittent lognormal distributions. Previous &egr; and &khgr; values are inferred from fossil parameters of microstructure such as vertical overturn scales of the density fluctuations, or from the Cox number. Primary active turbulent events start the mixing process, fossilize, and are eroded back toward the initial uniform-density-gradient nonturbulent state by parasitic secondary-active-turbulent events which complete the mixing. Comparisons are presented between laboratory, lake, ocean, and fjord data sets and the Gibson (1980-1987) fossil-turbulence model. ¿American Geophysical Union 1987