Observation-based parameterization of air-sea fluxes in terms of wind speed and atmospheric stability under low-to-moderate wind conditions

This study explores the behavior of the exchange coefficients for wind stress (C-D), sensible heat flux (C-H), and water vapor flux (C-E) as functions of surface wind speed (U-10) and atmospheric stability using direct turbulent flux measurements obtained from a platform equipped with fast-response turbulence sensors in a low-to-moderate wind region. Turbulent fluxes are calculated using the eddy-correlation method with extensive observations. The total numbers of quality-controlled 30 min flux runs are 12,240, 5813, and 5637 for estimation of C-D, C-H, and C-E, respectively. When adjusted to neutral stability using the Monin-Obukhov similarity theory (MOST), we found that C-DN, C-HN, and C-EN decrease with neutral-adjusted wind speed when wind speed is less than 5 m/s. C-DN is constant over the range 5 m/s<U-10N<12 m/s, then increases with U-10N when U-10N>12 m/s. In contrast, C-HN and C-EN exhibit no clear dependence on wind speed and are generally constant, with mean values of 0.96 x 10(-3) and 1.2 x 10(-3), respectively. This behavior of neutral exchange coefficients is consistent with the findings of previous studies. We also found that C-DN under offshore winds is generally greater than that under onshore wind conditions, which is ascribed to the younger wind waves present due to the shorter fetch in the former case. However, this behavior is not exhibited by C-HN or C-EN. The original C-D, C-H, and C-E values without MOST adjustment are also investigated to develop a new parameterization based on wind speed and stability. Three stability parameters are tested, including the bulk Richardson number, stability as defined in COARE 3.0, and a simplified Richardson number using the Charnock parameter. This new parameterization is free of MOST and the associated self-correlation. Compared with previous studies and COARE 3.0 results, the new parameterization using the simplified Richardson number performs well, with an increased correlation coefficient and reduction of root-mean-square error and bias.