EVALUATION OF CORRECTIONS ON TURBULENT FLUXES OBTAINED BY EDDY COVARIANCE METHOD IN HIGH WINDS

This paper investigates processing of fast-response data and corrections of turbulent fluxes obtained by using eddy covariance method based on data collected at an offshore observation tower during three Cold-intrusion (CI) events in the South China Sea in 2010. This study presents the data processing procedure in detail and compares frictional velocities (u[sub]·[/sub]), sensible heat fluxes (H) and latent heat fluxes (LE) yielded by using different averaging periods and different coordinate rotation methods; evaluates the sonic temperature correction for sensible heat flux and the Webb correction for latent heat flux as a function of 10 m wind speed (u[sub]10[/sub]) during the CIs. The results show (1) that the different averaging periods of 30 min and 10 min cause biases of u[sub]·[/sub] (H, LE) within 5% (15%, 62%). The values of u[sub]·[/sub] (H,LE) averaged from 30 mins are mostly larger than those averaged from 10 mins. We suggest that the averaging period of 10 min is not sufficiently long to capture all scale eddies and recommend 30 min averaging period in calculating turbulent fluxes using eddy covariance method during CIs; (2) that the values of u[sub]·[/sub] (H,LE) obtained from double rotation (DR2) and those obtained from planar fit rotation (PF) have good agreements and correlation coefficients between them are larger than 0.99. Because PF method requires unchanged environment and it is easier to apply DR2 method, we suggest DR2 coordinate rotation method in processing fast-response data; and (3) that the median values of frictional velocity (sensible heat flux and latent heat flux) binned according to 2 m s[sup]-1[/sup] intervals of u[sub]10[/sub] increase (decrease, increase) by less than 9% (4%, 10%) by Coriolis corrections (sonic temperature corrections, Webb corrections), which decreases (decreases, increases) with increasing u[sub]10[/sub] when u[sub]10[/sub] are 5-17 m s[sup]-1[/sup].