Comparisons of JEDI-Based 4DEnVar and 3DEnVar for Direct Radar Data Assimilation and Convection-Allowing Forecasts

High-frequency observations from weather radar can provide nearly continuous observations of weather systems. In practice, three-dimensional ensemble variational (3DEnVar) data assimilation (DA) usually groups frequent observations into small batches and performs the analyses through intermittent assimilation cycles. This approach needs frequent stopping and restarting of the prediction model, which can introduce shock to the prediction system when performing a new analysis every time. In addition, 3DEnVar assumes all observations taken over a chosen time window to be valid at the analysis time, which introduces large timing errors when the weather system is fast evolving. A better approach to more fully utilize observations collected over time is four-dimensional ensemble variational (4DEnVar), which can use observations distributed over time simultaneously and at the times when they are collected.

In this study, 3DEnVar and 4DEnVar within the Joint Effort for Data assimilation Integration (JEDI) framework are compared for the radar data assimilation and forecast in a storm case. In the preliminary test, radar reflectivity from the Multi-Radar Multi-Sensor (MRMS) reflectivity mosaic and radial wind from Weather Surveillance Doppler Radar (WSR)-88D level 2 BUFR data are assimilated by hourly 3DEnVar and 4DEnVar over the WoFS-like domain, respectively. In the case of 4DEnVar, the DA time window is configured by considering that the typical radar volume scans are about 5 minutes apart for WSR-88D, and the time information of each radar sweep is considered to construct 4D radial wind observation. 4DEnVar performs better than 3DEnVar in that 4DEnVar has better reflectivity analysis and precipitation forecasts in terms of fractions skill score for the hourly accumulated precipitation. We also conduct a series of sensitivity experiments to determine the optimal 4DEnVar configuration, such as the lengths of the DA time window and time-localization radii for reflectivity and radial velocity, respectively. More detailed results will be presented at the conference.