Evaluation of OAFlux datasets based on in situ air-sea flux tower observations over the Yongxing Islands in 2016

Abstract. The high-quality Yongxing air-sea flux tower (YXASFT), which was specially designed for air-sea boundary layer flux-related observations, was constructed on Yongxing Island in the South China Sea (SCS). Surface bulk variable measurements were collected during a one-year period from 2016/02/01 to 2017/01/31. The sensible heat flux (SHF) and latent heat flux (LHF) were further derived via the Coupled Ocean-Atmosphere Response Experiment version 3.0 (COARE3.0) using those variables. This study employed the YXASFT in situ observations to evaluate the Woods Hole Oceanographic Institute (WHOI) OAFlux reanalysis data products in the SCS. The study period was divided into the spring, summer_autumn and winter periods to conduct seasonal comparisons for each variable. First, the reliability of COARE3.0 data in the SCS was validated using direct turbulent heat flux measurements via an eddy covariance flux (ECF) system. The LHF data derived from COARE3.0 are highly consistent with the ECF measurements with a coefficient of determination (R2) of 0.78. Second, to conduct seasonal comparisons, the overall reliabilities of the bulk OAFlux variables diminish in order from Ta, U, Qa to Ts based on a combination of R2 values and biases. OAFlux overestimates (underestimates) U (Qa) throughout the year and provides better estimates of both variables in the winter and spring than in the summer_autumn period, which seems to be highly correlated with the monsoon climate in the SCS. The lowest R2 value is observed between the OAFlux-estimated and YXASFT-observed Ts, indicating that Ts is the least reliable product and should thus be used with considerable caution. In terms of the heat fluxes, OAFlux considerably overestimates LHF with an ocean heat loss bias of 52 w/m2 (73 % of the observed mean) in the spring, and the seasonal OAFlux LHF performance is consistent with U and Qa. The OAFlux-estimated SHF appears to be poorly representative with enormous overestimations in the spring and winter, while its performance is much better during the summer_autumn period. Third, an analysis reveals that the biases in Qa are the most dominant factor on the LHF biases in the spring and winter and that the biases in both Qa and U are responsible for controlling the biases in LHF during the summer_autumn period. The biases in Ts are responsible for controlling the SHF biases, and the effects of biases in Ts on the biases in SHF during the spring and winter are much greater than that in the summer_autumn period.