The application of superposition of Gaussian components in GRAPES-RAFS

The background error horizontal correlation model decides the structure of analysis increments affected by observations, and the analysis increments information changes with different scales in spectral space. The characteristics of Gaussian function (Gauss) and superposition of Gaussian components (SupG) are studied. Meanwhile, according to the structures of background error horizontal correlations estimated by NMC method, the correlation model is improved based on statistical results and characteristics of functions. The application of SupG correlation model can not only increase the meso- and small-scale information of analysis increments, but also mitigate the inappropriate large negative correlation because the correlation structure of wind with SupG is closer to actual weather. The analysis and forecast qualities between the control experiments (Gaussion function) and SupG experiments run with GRAPES-RAFS system are compared. The numerical results indicate that the convergence of the objective function is accelerated, and meso- and small-scale power spectra of analysis increments are increased so there are more meso- and small-scale information in analysis increments. The analysis qualities are obviously improved especially in wind analysis. Meanwhile the forecast qualities of 12 h wind are improved too. In addition, the precipitation forecast score ETS is higher in SupG experiments than in control experiments.