Test of a Hybrid 3DEnVAR and WRF-DART Analysis and Forecast System during the HWT Spring Experiments in 2017

A prototype system for the Warn-on-Forecast (WoF) project at NOAA, known as the NSSL Experimental Warn-on-Forecast System for ensembles (NEWS-e) has performed real-time experiments during the NOAA Hazardous Weather Testbed (HWT) several years since 2015 with promising results. Recently, a real-time hybrid analysis and forecast system based on 3DVAR ensemble analysis technique, the WRF-ARW forecast model and the NEWS-e system was developed. The purpose is to provide deterministic gridded analysis and forecast products to forecasters besides the probabilistic forecasts provided with the NEWS-e system. The deterministic analysis is done with the locally developed 3DEnVAR system and the forecast is performed with the WRF-ARW forecast model. The 3DEnVAR system incorporates available radar data, satellite retrieved cloud water path, and traditional soundings, and the 36 ensemble members from the NEWS-e forecasts provides flow-dependent cross-covariance for variables at each 15-minute data assimilation cycle. This system can ingest ensemble forecasts with dual-resolution capability, i.e. the WRF-DART analysis is performed at 3 km resolution and the 3DEnVAR analysis can be done at either 3 km or 1.5 km resolutions. Then, one deterministic 3 hour WRF forecast is launched every 30 minutes from the physically-consistent gridded analysis at high resolution.

This enhanced system has been tested during the 2017 HWT spring experiments period. Several configurations of experiments will be conducted. One set of experiments use the 3DVAR data assimilation analysis and forecast cycles only, which will be compared with the hybrid method mentioned above. The expectation is that the hybrid system should outperform the 3DVAR-only system because the 3DVAR-only method does not use the flow-dependent error covariance provided through the ensemble members. Another set of experiments will evaluate the impact of dual-resolution analysis cycles by comparing with the single-resolution analysis and forecast cycles. The detailed performance of this system and some preliminary results will be reported in this symposium.