Search

Surface latent heat fluxes help maintain tropical intraseasonal precipitation. We develop a latent heat flux diagnostic that depicts how latent heat fluxes vary with the near‐surface specific humidity vertical gradient (Δq) and surface wind speed (|V|). Compared to fluxes estimated from |V| and Δq measured at tropical moorings and the Coupled Ocean Atmosphere Response Experiment 3.0 (COARE3.0) algorithm, tropical latent heat fluxes in the National Center for Atmospheric Research CEMS2 and Department of Energy E3SMv1 models are significantly overestimated at |V| and Δq extrema. Madden–Julian oscillation (MJO) sensitivity to surface flux algorithm is tested with offline and inline flux corrections. The offline correction adjusts model output fluxes toward mooring‐estimated fluxes; the inline correction replaces the original bulk flux algorithm with the COARE3.0 algorithm in atmosphere‐only simulations of each model. Both corrections indicate reduced latent heat flux feedback to intraseasonal precipitation, in better agreement with observations, suggesting that model‐simulated fluxes are overly supportive for maintaining MJO convection. Plain Language Summary: Surface latent heat flux from ocean to the atmosphere is one of the important processes that provides water vapor and energy to the daily tropical rainfall. In this study, a visually intuitive latent heat flux diagnostic is proposed to better understand the model shortfall on its latent heat flux representation. This diagnostic allows a simple assessment of model latent heat flux biases arising either from biases in water vapor or surface wind speed as well as other empirical coefficients in the model. We demonstrate that, compared to "observed" fluxes also estimated from water vapor and surface wind speed measured at tropical moorings, tropical latent heat fluxes in the National Center for Atmospheric Research CEMS2 and Department of Energy E3SMv1 models are significantly overestimated when extreme water vapor or surface wind speed happens. Both offline and inline latent heat flux correction is applied to simulated fluxes. For both models, the correction reduces the percentage of latent heat flux on supporting the rainfall over the tropics which is in better agreement with observations. Particularly, the latent heat flux correction are nonuniform across different stages of the Madden–Julian oscillation (MJO). This finding suggests that a model improvement on the latent heat flux representation will change the simulated MJO. Key Points: Latent heat flux shows significant differences between bulk flux schemes as a function of wind speed and humidity disequilibriumChanging bulk flux schemes in model shows a more realistic latent heat flux contribution on maintaining precipitationLatent heat flux difference due to bulk flux schemes are nonuniform during different Madden–Julian oscillation phases [ABSTRACT FROM AUTHOR]