To understand these errors in HadGEM, inexpensive initialised 30-day hindcases of strong MJO events were performed at climate resolution (150 km). Many modifications were then applied, singly, to sub-grid parameterization schemes, spanning a range of likely causes of a deficient MJO. Two modifications improved the MJO: increasing the convective entrainment rate and switching off the convective momentum transport (CMT). These modifications were then applied to 20-year atmosphere-only simulations. Increasing the entrainment rate improved mid-tropospheric dry biases, suppressed spurious deep convection in dry columns, and systematically improved MJO amplitude and period. Removing the CMT gave minor improvements to the MJO but considerably degraded the mean state.
The high- and low-entrainment versions of HadGEM3 were then coupled to the KPP boundary-layer ocean model, using one-metre resolution in the mixed layer and three-hourly coupling. Introducing this coupling to a previous version of the Hadley Centre atmospheric greatly improved its MJO, but that version of the atmospheric model lacked a strong MJO. This presentation will address whether air-sea coupling brings further benefits to an atmosphere-only model that already produces a reasonable MJO.