Investigating A WRF-BASED Mixed Variational and Nudging Data Assimilation Scheme For The Us Army Convective-scale Nowcasting System

Abstract

The US Army Research Laboratory (ARL) has been developing and testing a WRF-based observation nudging FDDA (four-dimensional data assimilation) system to provide high spatial resolution ( ~1 km) and frequently updated (~ 1 hr) short range forecasts or ?nowcasts? of the battlefield environment out to the 3-6 hr time frame. This system is referred to by ARL as the Weather Running Estimate-Nowcast (WRE-N) and has the capability of assimilating asynoptic and sporadic direct observations from various sources such as local soundings, surface observations, and aircraft in-situ measurements. The FDDA nudging technique used by the WRE-N has been proven in numerous past research studies to be both efficient and effective at such convection-resolving scales; however, it lacks the capability to assimilate indirect observations such as radar radial wind/reflectivity and satellite radiance data. In order to take advantage of indirect local observations from various remote sensing instruments (often of resolution critical to convection-resolving scales), a mixed or ?hybrid? data assimilation system has been developed and tested for the WRE-N model. This technique uses the observation or ?station? nudging for direct observations in addition to three-dimensional (3D) grid nudging using analyses created by a different assimilation technique in order to combine the advantages of both nudging and 3DVAR. The 3D analyses used for the grid nudging is produced by the Variational LAPS (Local Analysis and Prediction System) developed by National Oceanic and Atmospheric Administration (NOAA) Earth System Research Laboratory's Global System Division (GSD). An added advantage of the Variational LAPS is its capability to process both direct and indirect observations from many sources such as radar, satellite, lidar, wind profiler, and GPS, among others in its multi-scale analysis. The multi-scale variational analysis created by variational LAPS is also used as the WRE-N initial conditions in order to ?hotstart? (more rapidly spin up) the WRE-N model. The effectiveness of utilizing the Variational LAPS analysis for analysis nudging in combination with standard observation nudging is demonstrated initially through well-selected case studies. A set of numerical experiments to test the mixed data assimilation scheme was carefully designed and carried out. Results from two cases, one a tornadic supercell event from Moore, OK, in 2013, the other a winter storm event over the eastern US in 2014, will be presented. It is expected that the WRE-N model utilizing both observation nudging and analysis nudging using the Variational LAPS analysis (assimilating radar and satellite data) will improve forecasting skill and predictability, especially in the 0-3 hr time frame.