Assessing surface heat fluxes in atmospheric reanalyses with a decade of data from the NOAA Kuroshio Extension Observatory

Abstract

Listen

Translate article

AbstractPrevious studies have found large biases and uncertainties in the air‐sea fluxes from Numerical Weather Prediction model reanalyses, which must be identified and reduced in order to make progress on weather and climate predictions. Here air‐sea heat fluxes from NOAA Kuroshio Extension Observatory (KEO) measurements are used to assess two new reanalyses, NCEP's Climate Forecast System Reanalysis (CFSR) and ECMWF Reanalysis‐Interim (ERA‐I), suggesting that these two new generation reanalyses have significantly improved. In both reanalyses, all four flux components (sensible and latent heat flux and net longwave and shortwave radiation) are highly correlated with observation, with the correlation of total net surface heat fluxes above 0.96. Although errors of the net surface heat flux have significantly reduced from previous reanalyses, the Root Mean Square Errors (RMSEs) and biases remain high especially for CFSR: the RMSEs of CFSR and ERA‐I are reduced by 25–30% to 64 and 61 W/m2, respectively, while biases are reduced by 40–60% to 28 and 20 W/m2. But CFSR overestimates the winter heat release by 90 W/m2. The main cause of biases is the latent heat flux, while RMS errors are primarily due to latent heat flux and shortwave radiation errors. Both reanalyses overestimate the wind speed associated with winter storms and underestimate specific humidity in summer. The ERA‐I latent heat flux and its total net surface heat flux are however closer to observation. It is the bulk algorithm in CFSR that is found to be mainly responsible for overestimates of winter heat release in CFSR.