Abstract The sensitivity of the ECMWF analysis-forecast system for four tropical cyclones during the period spanned by the Australian Monsoon Experiment to cumulus parameterization and model resolution is presented. Two parameterization schemes were compared: namely, the Kuo cumulus parameterization and the Betts-Miller adjustment scheme. Both analyses and forecasts show considerable sensitivity with the Betts-Miller scheme generating more intense cyclonic systems as indicated by maps of sea-level pressure, low-level winds, vorticities, and cross sections in the neighborhood of the cyclones. The Betts-Miller scheme also results in better consistency of the fields in the vertical between the divergent circulation and vertical velocity. Of the two cases considered, one (for tropical cyclone Jason) showed a marked sensitivity to increased model resolution with the higher resolution forecast showing considerable improvement.