Design and Implementation of a GSI-based Convection-Permitting Ensemble Data Assimilation and Forecast System for PECAN: Impacts of model physics, resolution, and assimilation of PECAN observations

A real-time GSI-based and ensemble-based data assimilation (DA) and forecast system was implemented at the University of Oklahoma for PECAN.  This talk will summarize the design of the system, including experiments conducted using retrospective forecasts from 2014, and give an overview of the system performance for forecasting nocturnal convection during PECAN.

Retrospective forecasts from 2014 showed that the impacts of radar DA between 1200 and 1300 UTC, as well as the frequency and number of DA cycles and DA physics configuration, extend through the following night. The Thompson microphysics scheme for DA better analyzed the effects of morning convection on environmental moisture than WSM6, which improved the convection forecast the following night.  A multi-PBL configuration during DA also led to less skillful short-term forecasts than even a relatively poorly performing single-PBL scheme.  During the 2014 retrospective forecasts, Thompson microphysics and MYNN PBL provided the most skillful nocturnal precipitation forecasts.

During the 2015 PECAN field experiment, a 20-member 4km ensemble forecast was initialized daily at both 1300 and 1900 UTC, together with a 1km deterministic forecast at 1300 UTC.  Subjective and objective evaluation of the real-time forecasts from 1 June through 15 July 2015 was conducted, with an emphasis on nocturnal Mesoscale Convective Systems (MCSs), nocturnal Convective Initiation (CI), nocturnal Low Level Jets (LLJs) and bores on the nocturnal stable layer. For example, LLJ and bore forecasts demonstrated sensitivity to PBL scheme and horizontal resolution, respectively.  Verification of nocturnal precipitation during overnight hours, a proxy for MCSs, showed both greater skill and spread for the 1300 UTC forecasts than the 1900 UTC forecasts.  Objective verification of forecast CI timing also revealed strong sensitivity in forecast CI timing to PBL scheme but an overall unbiased ensemble. 

A case study of a comprehensively observed MCS/cold pool/bore system on 11 July 2015  during a PECAN IOP is also used to investigate the impacts of model resolution and assimilation of PECAN observations for improving predictability of such nocturnal convective systems.