Choice of Bulk Surface Flux Algorithm in Models May Affect MJO Simulation Skill
Previous work has shown that fluxes of water vapor and heat between the atmosphere and ocean are overestimated by 10%-20% for many algorithms used in climate models. Using a new method to diagnose surface fluxes, we analyzed conditions where these biases are most problematic for maintaining and propagating tropical convection within the Madden-Julian oscillation (MJO). Two approaches demonstrated how improved representation of surface fluxes in climate models could lead to improved MJO simulation.
The MJO regulates weather patterns and precipitation extremes throughout the tropics as well as globally through its influence on tropical-extratropical teleconnections. Biases in surface fluxes that may vary throughout the MJO lifecycle may lead to erroneous maintenance and propagation of MJO convection that can negatively affect MJO simulation and its teleconnections. While a substantial body of previous work has shown that the sensitivity of tropical convection to mid-level moisture is an important process for the MJO, our results indicate that interactions of convection with surface fluxes may also affect MJO simulation.
Surface fluxes computed with models' default surface flux algorithms are overly supportive of maintaining MJO convection and under-supportive of propagating MJO convection. We examined the effect of surface flux biases during the MJO lifecycle using two different methods. First, using output from historical simulations of the E3SMv1 and CESM2 models, we estimated corrected surface fluxes by adjusting model fluxes to those that would have been calculated using the COARE3.0 bulk flux algorithm (Fairall et al. 1996; 2003). This "offline" correction suggested that running the models with the COARE algorithm in place of the default algorithm should lead to reduced surface flux support for MJO maintenance, and improved surface flux support for MJO propagation. We then performed "inline" surface flux corrections by running atmosphere-only versions of both models with the COARE3.0 bulk flux algorithm. While uncoupled simulations of both models produce westward- instead of eastward-propagating intraseasonal convection (a trait shared by many other models in uncoupled configurations), this bias was reduced in the experiments with COARE3.0 fluxes, in better agreement with observations. Our results indicate that subtle changes to surface flux magnitude throughout the MJO lifecycle can have important consequences for MJO simulation.