Valid Time Shifting for an Experimental RRFS Convection-Allowing EnVar Data Assimilation and Forecast System: Description and Systematic Evaluation in Real Time

Abstract

Abstract This study describes a real-time implementation of valid time shifting (VTS) within the Gridpoint Statistical Interpolation–based ensemble-variational (EnVar) data assimilation system, developed at the Multi-Scale Data Assimilation and Predictability Laboratory. This system, featuring data assimilation of mesoscale conventional observations and storm-scale radar reflectivity observations and interfaced with the next-generation Finite Volume Cubed Sphere Dynamical Core limited-area model (FV3-LAM), was run in real-time during the 2021 Hazardous Weather Testbed Spring Forecast Experiment. The VTS method efficiently increases ensemble size by incorporating ensemble forecast output before and after the central analysis. Two configurations were examined to systematically evaluate VTS: a baseline 36-member system with hourly data assimilation (NOVTS), and an experiment testing VTS for the radar analysis step. Verification across 22 cases shows statistically significant benefits of VTS to increase ensemble spread and better fit first guesses to observations. Control member forecasts launched at 0000 UTC have consistently higher skill, lower bias, and higher reliability in VTS than in NOVTS throughout the 18-h forecast evaluation period, especially from severe cases often featuring upscale growth into mesoscale convective systems. Verification of updraft helicity-based ensemble surrogate severe probabilistic forecasts against observed storm reports shows higher skill of VTS when verifying on finer scales, with benefits to constraining higher probabilities over report locations and reducing probabilities over no-report locations. This study is a first step toward the next-generation Rapid Refresh Forecast System (RRFS), demonstrating the feasibility of such a real-time system and the potential benefits of VTS implementation.