Premium
State estimation of convective storms with a two‐moment microphysics scheme and an ensemble Kalman filter: Experiments with simulated radar data
Author(s) -
Xue Ming,
Jung Youngsun,
Zhang Guifu
Publication year - 2010
Publication title -
quarterly journal of the royal meteorological society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.744
H-Index - 143
eISSN - 1477-870X
pISSN - 0035-9009
DOI - 10.1002/qj.593
Subject(s) - parametrization (atmospheric modeling) , data assimilation , moment (physics) , ensemble kalman filter , kalman filter , mathematics , physics , meteorology , statistical physics , statistics , extended kalman filter , quantum mechanics , radiative transfer
The ability of the ensemble Kalman filter method to estimate an increased number of state variables associated with a double‐moment (DM) microphysics scheme is examined for the first time through observing system simulation experiments, assuming either a perfect or imperfect prediction model and/or observation operators. With the DM scheme, mixing ratios and total number concentrations of hydrometeor species are predicted. It is found that the increased number of state variables can be reasonably well estimated when both radial velocity ( V r ) and reflectivity ( Z H ) observations are used and when the prediction model is assumed to be perfect. However, the errors increase significantly when Z H is used alone. In this case, the filter has difficulty in estimating independently‐varying mixing ratios and number concentrations, which are both directly involved in the calculation of Z H . The addition of V r data helps alleviate a problem associated with the solution not being sufficiently constrained by observations. With the DM scheme, the correlations between Z H and model state variables exhibit complex spatial structures that depend on the location of the Z H observation. Collocated Z H and vertical velocity show negative correlation when the observation is taken where ice phase hydrometeors are dominant, but positive correlation when it is taken where large quantities of liquid hydrometeors exist. Further study is needed to fully understand the complex correlation structures. Imperfect model experiments were performed, with two types of model errors: (1) microphysical parametrization error due to incorrectly assumed shape parameter of the gamma particle size distribution (PSD), and (2) different ways of calculating hydrometeor scattering. The results show that the model error degrades the state estimation in general. Nevertheless, the estimated states are still reasonably good when both V r and Z H are assimilated. Perturbing the shape parameter of gamma PSDs within the forecast ensemble improves the overall state estimation. Copyright © 2010 Royal Meteorological Society