Abstract:Graded straight-line wind warning is a challenging issue in refined short-term and nowcasting forecasting. Based on S-band polarimetric Doppler weather radar data and hourly maximum wind speed data, this study selected 40 straight-line wind events caused by isolated severe storms (echo intensity ≥60 dBZ) during spring 2024. These events were classified into strong straight-line wind events (wind force scales 9-13) and weak straight-line wind events (wind force scales ≤7) according to surface maximum wind force. The analysis reveals that the storm structure characteristics of strong straight-line wind events are similar to the supercell radar features that produce winds above force scale 10. The vast majority of storms in the strong wind group exhibit persistent mesocyclones and other supercell structural features, with their echo intensity, Vertically Integrated Liquid (VIL), mid-level radial convergence intensity, and storm motion speed significantly higher than those of the non-strong wind group. Based on statistical analysis, this study proposes strong wind warning indicators and lead times. Notably, the descent of reflectivity factor cores is a common feature of severe storms and cannot be used to distinguish straight-line wind intensity. The differential reflectivity trough (ZDR trough) is considered a polarization characteristic of strong straight-line winds. 40% of the strong wind group cases showed a ZDR trough 42 minutes before the strong wind occurred, indicating that the strong downdrafts induced by the phase changes (e.g., melting and evaporation) of hail or graupel particles during their descent represent one mechanism for straight-line winds production. A relatively continuous ZDR trough can serve as a precursor signature for such strong straight-line winds. The results of this study clearly identify the precursor signatures of isolated strong thunderstorms that cause damaging straight-line winds in the spring, providing a scientific basis for windstorm forecasting and the design of objective algorithms.