Hierarchical modelling of tropical convective systems using explicit and parametrized approaches

Abstract

Cloud systems observed during 1–7 September of GATE are examined in a hierarchical approach, namely: a two-dimensional cloud-resolving simulation using a 2 km grid length; two- and three-dimensional simulations using the Kain-Fritsch convective parametrization and 10, 15 and 25 km grid lengths; and coarse-grid simulations without any convective parametrization. All simulations are forced by the same objectively analysed time-varying large-scale advection of temperature and moisture. The domain-mean winds are relaxed to the observed wind profiles. Both the cloud-resolving modelling and the lower-resolution modelling with parametrized convection realize the three observed cloud system categories (squall line, non-squall cluster and scattered convection) and transitions among them. In particular, the well-organized fast-moving squall-type cloud system observed on 4 September is realized in a three-dimensional experiment with parametrized convection. In contrast, the lower-resolution modelling without any convective parametrization fails to produce the squall-type convective system during the weakly forced period but successfully represents the non-squall cluster during strong forcing. This lack of success is mostly attributed to convective triggering and the absence of resolved downdraught-enhanced surface fluxes. These issues are not as critical during strong large-scale forcing. The observed evolutions of temperature, water vapour mixing ratio, precipitation and surface moisture flux are realized in all simulations. A common deficiency is the overprediction of upper-level relative humidity. The simulation with parametrized convection features a comparatively large low-level water vapour mixing ratio, a surface and upper-level cold temperature bias and a mid-tropospheric warm bias. This is mainly attributed to deficiencies in how the Kain-Fritsch scheme represents convective mass flux, detrainment and entrainment by cumulus congestus.