The parameterization of the turbulence used in one-dimensional oceanic mixed layer models is briefly reviewed, focusing on the long-term response of these models. Particular attention is directed towards the parameterization of turbulent dissipation. A general parameterized form is proposed that provides a useful guideline to devise new parameterizations and to compare existing ones. Different models of the classical Niiler–Kraus type are first tested by simulating four years (1969–72) of the upper ocean evolution at Ocean Station Papa. In the results, distinction is made between the errors inherent in the model and those due to changes in the upper ocean heat content not explained by the surface heat fluxes. It appears that, after a needed empirical calibration, all models systematically overestimate the sea surface temperature (SST) in summer and underestimate it during fall. In absolute value, the maximum error on the monthly-mean predicted SST reaches about 1 K. In an attempt to reduce this e... Abstract The parameterization of the turbulence used in one-dimensional oceanic mixed layer models is briefly reviewed, focusing on the long-term response of these models. Particular attention is directed towards the parameterization of turbulent dissipation. A general parameterized form is proposed that provides a useful guideline to devise new parameterizations and to compare existing ones. Different models of the classical Niiler–Kraus type are first tested by simulating four years (1969–72) of the upper ocean evolution at Ocean Station Papa. In the results, distinction is made between the errors inherent in the model and those due to changes in the upper ocean heat content not explained by the surface heat fluxes. It appears that, after a needed empirical calibration, all models systematically overestimate the sea surface temperature (SST) in summer and underestimate it during fall. In absolute value, the maximum error on the monthly-mean predicted SST reaches about 1 K. In an attempt to reduce this e...