Improvement of finescale parameterization through reducing uncertainty in spectrum estimation

Abstract

Finescale parameterization (FP) is employed widely to estimate the large-scale distribution of internal wave-induced mixing, which is crucially important for the development of ocean general circulation models. In this study, FP performance was evaluated using hydrographic and microstructure measurements extracted from the Microstructure Program dataset. A general tendency of overestimation with increase in the estimated internal wave energy level was observed. Using the Monte Carlo method, the turbulent dissipation rates under prescribed spectra were estimated to illustrate how uncertainty in spectrum estimation contributes to the bias. The overestimation tendency was replicated under the FP by the commonly used periodogram spectral method. By replacing the periodogram method with an autoregressive (AR) spectral estimator, the overestimation tendency was reduced considerably. Application of FP with the AR method to the collected hydrographic data greatly reduced the bias, with the root mean square error reducing from 0.72 to 0.46, the variance of the bias decreasing from 0.57 to 0.23, and the correlation of the bias with the internal wave energy level reducing from 0.62 to 0.32, in base-10 logarithmic coordinates. Application of FP with the AR spectrum estimator would help in estimating diapycnal mixing within the ocean interior more accurately and increase the robustness of FP.