Premium
Efficient moist physics schemes for data assimilation. II: Deep convection
Author(s) -
Stiller O.
Publication year - 2009
Publication title -
quarterly journal of the royal meteorological society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.744
H-Index - 143
eISSN - 1477-870X
pISSN - 0035-9009
DOI - 10.1002/qj.362
Subject(s) - nonlinear system , parametrization (atmospheric modeling) , mathematics , convection , meteorology , linearization , data assimilation , stability (learning theory) , statistical physics , computer science , physics , quantum mechanics , radiative transfer , machine learning
The development of a cost‐effective convection scheme for a linear 4D‐Var model is described. The very noisy and unsteady behaviour of the Met Office's nonlinear mass‐flux convection scheme motivated an approach in which the linear model is designed to correspond to its nonlinear counterpart only in a time‐average sense. Technically this is implemented in a numerically efficient way by pre‐processing (mainly time‐averaging) the linear scheme's main input files outside the linear model. The new linear scheme includes only two terms from the nonlinear parametrization's sensitivities; one originates from the subsidence part, while the other, the dominant term, involves convective mass‐flux perturbations resulting from the convective available potential energy (CAPE) closure relationship. Including these terms required error mitigation measures, partially to compensate for neglected terms that in some situations have the opposite sign to the represented terms. Measures were also taken to increase the scheme's stability as well as its numerical efficiency. The introduction of this scheme into the Met Office numerical weather prediction (NWP) system led to clear improvements of the forecast skill. Linearization test results show that the new scheme improves the match between linear and nonlinear models. Consistent with its statistical interpretation, precipitation increments of the linear scheme show a reasonable correlation to their nonlinear counterparts only if the nonlinear precipitation is averaged over time intervals of several hours. ©Crown Copyright 2009. Reproduced with the permission of HMSO. Published by John Wiley & Sons Ltd.