Evaluation of multi-source WRF data assimilation for a surface-reaching tropical hail event in Yogyakarta, Indonesia: a case study

Abstract

Surface-reaching hail in the humid tropics is rare because hailstones must survive descent through a high freezing level and deep warm layer. This study evaluates how multi-source WRFDA-3DVAR data assimilation affects the representation of tropical convection that produced surface-reaching hail over Yogyakarta, Indonesia, on 11 March 2025. Surface reports, C-band radar, Himawari-9, radiosonde observations, and ERA5 diagnostics were used to reconstruct the event and provide an observational reference. Seven convection-permitting WRF experiments were conducted, including a no-assimilation control and configurations assimilating conventional observations, radar reflectivity, radar radial velocity, satellite observations, and their combinations. A hydrometeor-based hail-supporting proxy was used as a supplementary mixed-phase diagnostic, separate from standard WRF diagnostic reflectivity. Observations showed rapid convective intensification between 08:00 and 08:20 UTC, coinciding with surface hail near 08:15 UTC. Himawari-9 showed cloud-top cooling from approximately − 53 °C to − 72 °C within 20 min, while radar indicated vertical deepening above the ~ 5-km freezing level. Radar reflectivity assimilation produced the lowest aggregate SAL distance and centroid-distance error, indicating the most favorable storm-object morphology and placement. However, the no-assimilation control achieved the most balanced overall multi-metric performance, whereas all-source assimilation increased the hit rate but produced excessive spatial expansion. These results show that the benefit of multi-source data assimilation is strongly observation- and metric-dependent, highlighting the value of observation-specific WRFDA experiments for evaluating compact tropical hail-producing convection.