Jiafen Hu - November 11

Jiafen Hu Test of Different Transformation Functions to Hydrometeor and Water Vapor Mixing Ratios for Direct Variational Assimilation of Radar Reflectivity Data Location: https://meet.google.com/ksh-txvg-kni Date:     2020/11/20 Time:     3:00 PM Series:   Convective Meteorology (Mesoscale Dynamics) Abstract: There are many issues arose from using highly nonlinear radar reflectivity forward observation operator in three-dimensional

Start

November 20, 2020 - 3:00 pm

End

November 20, 2020 - 4:00 pm

Jiafen Hu

Test of Different Transformation Functions to Hydrometeor and Water Vapor Mixing Ratios for Direct Variational Assimilation of Radar Reflectivity Data

Location: https://meet.google.com/ksh-txvg-kni

Date:     2020/11/20

Time:     3:00 PM

Series:   Convective Meteorology (Mesoscale Dynamics)

Abstract: There are many issues arose from using highly nonlinear radar reflectivity forward observation operator in three-dimensional variational data assimilation methods (3DVAR), especially with hydrometeor mixing ratios as control variables (denoted as Q). When hydrometeor mixing ratios from model background are very small, the cost function gradient can be extremely large which causes slow convergence. In order to solve the problem, two methods were proposed recently. One uses logarithmic hydrometeor mixing ratios (LOGQ) as control variables during minimization process. Another uses power transformed mixing ratios (PQ), which applies a power parameter p to the variable transformation, as new control variables. In this study, all three methods are implemented in a weather adaptive, high-resolution, deterministic Warn-on-Forecast analysis and forecast system and tested on two severe weather events that occurred in May 2019. Radar reflectivity and radial velocity are assimilated. B
oth qualitive and quantitative evaluation are performed on 0-3h forecasts launched hourly from 1900 to 2300 UTC for each of the 2 cases. It is found that reflectivity analysis in experiments with PQ and LOGQ as control variables are better than those experiments with Q. The convergence of cost function minimization with PQ and LOGQ is significantly faster than the experiments with Q. It is also found the nonlinear transformation in LOGQ method can produce spurious analysis increments sometimes. Using PQ as control variables can alleviate this problem which produces less spurious analysis increment and improves analysis of composite reflectivity compared to Q and LOGQ as control variables.