Enhancing High-Impact Rainfall Forecasts via Multi-Sensor, Multi-Stage Variational Data Assimilation in WRF: Insights from the CYGMEN Project

<jats:p>Assimilating high-resolution radar reflectivity and complementary observation types remains a central challenge for improving short-term quantitative precipitation forecasts of severe Mediterranean weather systems. This study investigates the effect of multi-source data assimilation configurations using the Weather Research and Forecasting (WRF) model and its three-dimensional variational data assimilation system (3D-Var) to simulate the impact of storm Bora over Greece. Specifically, we evaluate the adjustments induced by synchronous versus multi-stage sequential assimilation of radar reflectivity, Global Navigation Satellite System (GNSS) precipitable water vapor, and surface synoptic observations (SYNOP), while exploring the impact of single versus multiple radar cost-function minimization loops. Diagnostic analysis of thermodynamic increments and low-level transport, combined with categorical forecast verification metrics, reveals that two-stage sequential configurations systematically outperform synchronous methods across light to heavy precipitation regimes. Synchronous ingestion induces localized diabatic heating that artificially deforms the 850 hPa geopotential height field, displacing low-level thermodynamic pathways westward. Conversely, sequential configurations introduce radar echoes into a moisture-corrected background state, preserving favorable thermodynamic profiles and localized storm trajectories. Furthermore, a distinct trade-off governs radar minimization: three minimization loops refine widespread stratiform structures (15–20 mm), whereas a single loop prevents the over-smoothing of peak intensities, yielding enhanced skill for heavy convective cores (30–40 mm). Finally, anchoring the lower troposphere with SYNOP observations provides a horizontal constraint that restricts low-level moisture transport, yielding a substantial relative improvement in predicting short-term heavy rainfall. These insights provide an operational framework for optimizing multi-source data assimilation setups during severe convective events in complex marine-continental interfaces.</jats:p>