A convection scheme for data assimilation: Description and initial tests

A new simplified parametrization of subgrid-scale convective processes has been developed and tested in the framework of the ECMWF Integrated Forecasting System for the purpose of variational data assimilation, singular vector calculations and adjoint sensitivity experiments. Its formulation is based on the full nonlinear convection scheme used in ECMWF forecasts, but a set of simplifications has been applied to substantially improve its linear behaviour. These include the specification of a single closure assumption based on convective available potential energy, the uncoupling of the equations for the convective mass flux and updraught characteristics and a unified formulation of the entrainment and detrainment rates. Simplified representations of downdraughts and momentum transport are also included in the new scheme. Despite these simplifications, the forecasting ability of the new convective parametrization is shown to remain satisfactory even in seasonal integrations. A detailed study of its Jacobians and the validity of the linear hypothesis is presented. The new scheme is also tested in combination with the new simplified parametrization of large-scale clouds and precipitation recently developed at ECMWF. In contrast with the simplified convective parametrization currently used in ECMWF's operational 4D-Var, its tangent-linear and adjoint versions account for perturbations of all convective quantities including convective mass flux, updraught characteristics and precipitation fluxes. Therefore the new scheme is expected to be beneficial when combined with radiative calculations that are directly affected by condensation and precipitation. Examples are presented of applications of the new moist physics in 1D-Var retrievals using microwave brightness temperature measurements and in adjoint sensitivity experiments.