Real Time Multi-scale GSI-based ensemble data assimilation and forecasting using radar and in-situ observations in support of the 2015 PECAN field campaign

The Plains Elevated Convection At Night (PECAN) field experiment (1 June – 15 July 2015) was designed to collect unprecedented in situ and remotely sensed observations of nocturnal convection and related phenomena. Specific foci of PECAN include nocturnal Mesoscale Convective Systems (MCSs), nocturnal Convective Initiation (CI), atmospheric bores and the nocturnal Low Level Jet (LLJ). A multi-scale GSI-based Ensemble Kalman Filter (EnKF) data assimilation (Johnson et al. 2015) and WRF-based ensemble forecast system was configured and implemented in real time during PECAN to support these specific and unique foci. The data assimilation component of the system includes the assimilation of synoptic scale surface and rawinsonde observations as well as NEXRAD radar reflectivity and velocity observations using a 40-member EnKF. The forecast component of the system includes a 20 (10) member ensemble forecast out to 24 (48) hour lead time over a 1200 x 1380 km domain covering the central United States with 4 km grid spacing, as well as a deterministic forecast at 1 km grid spacing over a similar domain. The forecasts are initialized daily at both 1200 and 1800 UTC.

This data assimilation and forecast system provided PECAN-specific guidance to forecasters and nowcasters for planning the deployment of the mobile observation platforms during the PECAN field phase. The system also provides a modelling framework that will be used to improve the understanding and prediction of nocturnal convection using the observations collected during PECAN. The real-time assimilation of NEXRAD data on a convection-permitting grid using EnKF is an important component of the data assimilation and forecast system.

While implementing the real-time system, experiments were conducted to understand the impacts of the number of radar data assimilation cycles, the frequency of radar data assimilation, multi-physics and multi-parameter ensemble configurations, and the impact on the ensemble forecasts of the radar component of data assimilation, among other system details. Initial results during pre-implementation experiments have shown that the relative advantage of assimilating radar observations at 5, rather than 10, minute intervals depends on the features of interest. Advantages of the more frequent assimilation are seen at weaker reflectivity thresholds while the less frequent assimilation is more advantageous for stronger convection during the early forecast hours. Initial experiments have also shown better analyses with the WSM6 microphysics scheme for the data assimilation component of the system, rather than Thompson microphysics. However, better performance of the forecast component was found with the Thompson scheme, rather than WSM6. The WSM6 scheme for data assimilation is further improved through the use of parameter perturbations. These results and others from these experiments will be presented together with verification of the overall real-time system performance, and modelling results obtained from the observations collected during PECAN.