Investigation of stochastic effect in an advanced global atmospheric model to improve weather and climate prediction

This study attempts to establish an efficient spectral dynamical core using double Fourier series (DFS) as a basis function, and to apply stochastic approach to dynamical process to improve weather and climate predictability. The DFS spectral model is quantitatively assessed on various time scales over the globe as compared to the traditional spherical harmonics (SPH) model. For vertical discretization, a sigma-pressure hybrid vertical coordinates is newly implemented into the DFS dynamical core, and moisture advection has been replaced with a mass-conserving semi-Lagrangian advection scheme to avoid negative values introduced by spectral transform. To represent model uncertainty, a stochastic representation of random error associated with dynamical processes (i.e., stochastic dynamics) is introduced, and its impact on medium-range forecast is discussed. Model random errors are simulated by multiplying nonlinear tendencies on grid space by a random number sampled uniformly from different intervals. In addition, time- and layer-dependent stochastic forcing is taken into consideration. Further details will be discussed in the conference.