Assimilation of GOES-16 Satellite Geostationary Lightning Mapper Lightning Flash Rate Data for the Analysis and Forecasting of Convective Storms Using EnKF and En3DVar Hybrid Methods (Invited Presentation)

The GOES R-Series satellites launched into space in the past few years carry the Geostationary Lighting Mapper (GLM), which measures the total lightning flash rates. The data are available in the form of lighting flash extent density (FED) data products at roughly 8 km pixel resolution and 20 second time intervals over the continental United States. The GLM data can complement ground-based weather radar data, and fill gaps left by the operational radars, which have poor coverage over mountainous regions, oceans, and can have gaps due to radar outages. Past studies on lighting data assimilation (DA) have shown the potential of achieving similar impacts on convective storm analysis and prediction as ground-based radar data. Past studies with FED data were either based on OSSEs or using pseudo FED data, however.

Capabilities are developed to assimilate GOES-R GLM FED data within the operational GSI framework, using ensemble Kalman filter (EnKF), and ensemble-3DVar (En3DVar) hybrid methods. FED observation operators based on graupel mass (FEDM) or graupel volume (FEDV) are used in the EnKF and while FEDM is used in the variational framework. The capabilities are tested with a mesoscale convective system (MCS) case and a case with multiple supercells. When tested with the MCS case with EnKF, FEDM and FEDV observation operators are found to perform similarly, and the assimilation of FED data shows clear positive impact. FED DA is effective in identifying and properly initializing regions of intense convection and helps improve the forecasts up to 3 hours. Direct adjustment to graupel as well as to other model states (such as vertical velocity) through flow-dependent cross-covariances work together to produce more consistent analyses then improved forecasts. 3DVar, EnKF and En3DVar are applied to the supercell case, and FED DA again leads to positive impacts on analyses and forecasts of the storms. EnKF and hybrid En3DVar are found to perform similarly while 3DVar performs quite a bit worse. Overall, the 2D FED data show great promise for initializing convective storms via advanced DA methods, especially in the absence of 3D volumetric radar data.