On Non-Gaussian Probability Densities on Convection Initiation and Development using a Particle Filter with a Storm-Scale Numerical Weather Prediction Model
Non-Gaussian probability densities in convection initiation (CI) and development are investigated using a particle filter with a numerical weather prediction model (NHM-PF). An observation system simulation experiment (OSSE) is conducted with a storm scale of 2-km grid spacing and 36 of observations and 1,000 of particles. The observations are created from a nature run, which simulates a well-developed cumulonimbus. For evaluation of non-Gaussianity, we propose to apply the Bayesian Information Criterion to compare the goodness of fit of three statistical models of Gaussian, two-Gaussian mixture and histogram. The PDFs become strongly non-Gaussian, when NHM-PF produces diverse particles over the CI period. This is led by non-Gaussian PDF of updraft at the beginning, and then the upper-bounded PDF of relative humidity, which creates non-Gaussian PDFs of QV and PT. The PDFs of cloud water and QR are quite far from Gaussian distributions throughout the experimental period. From these examination in addition to examinations on ensemble mean and spreads, it is concluded that the source of non-Gaussian in the CI is updraft.
Dr. Takuya Kawabata Meteorological Research Institute