Using multipliers that statistically tag errors arising from individual terms of model physics and dynamics in a forecast model for reduction of systematic errors

We present a statistical approach to estimate the significance of individual components of the physical parameterization and dynamical core in general circulation models. In our study, we used this approach for two global spectral models to not just estimate the significance of individual components but also reduce the systematical errors of these models. In our stochastic approach, we first estimate the contributions of the various constituting tendencies of the prognostic equations of the model; later on these model tendencies are regressed against the observed total tendency using multilinear regression technique to get three dimensional fields of multipliers for each component tendency. These constituting tendencies are then corrected by applying their corresponding field of multipliers. In turn, any deviation of a component tendency from its corrected version highlights the error of the model related to that particular component. These error estimates provide information on geographical and vertical distribution of forecast errors contributed by features such as nonlinear advective dynamics, deep cumulus convection, large-scale condensation physics, radiative processes, and the rest of physics and dynamics. Moreover, we go one step ahead and plug in these corrected tendencies in our models to reduce the systematic errors arising due to aforementioned components.