Implementation of Cloud-related Non-gaussian Background Error Statistics In Hybrid Data Assimilation For Convective-scale Prediction

Abstract

Current data assimilation for numerical weather forecasting at several NOAA operational centers is based on Bayesian statistics and probability, which are inherently tied to the Gaussian assumption of the probability density function of the observation and background errors. However, in reality, the errors involved in key variables linked to clouds and precipitation processes exhibit probability density function distributions that are far from being Gaussian. This is an important challenge to convective-scale forecasting of severe weather events, as these are strongly tied to clouds and precipitation. With the goal of improving the performance of operational data assimilation and convective-scale prediction, here we present the initial steps to augment the capabilities of the NOAA/NCEP Gridpoint Statistical Interpolation-hybrid system to account for the non-Gaussianity in the background error distributions of cloud hydrometeors for the Rapid Refresh and the High Resolution Rapid Refresh models.