Improving the prediction of the 3 July 2019 Kaiyuan tornadic supercell and embedded tornado with radar and surface data assimilation

The 3 July 2019 Kaiyuan City, Liaoning Province, tornadic supercell is simulated with the Weather Research and Forecasting (WRF) model, using up to five levels of nested grids, with the innermost grid having a 50‐m grid spacing. Reflectivity and radial velocity data from seven operational Doppler weather radars in different wavebands are assimilated together with dense automatic weather station (AWS) data using a 3DVAR data assimilation (DA) and cloud analysis system on the third‐level 450‐m grid with five‐minute cycles. Experiments assimilating surface data in a different number of cycles show that the analyses and forecasts are improved in terms of the intensity and structures of the simulated supercell and cold pool, compared to the one without assimilating surface data. The frequent assimilation of high‐resolution surface data tends to produce spurious convection due to imbalances introduced, which has a negative impact on the development of near‐surface rotation. With surface data assimilated in more cycles, more convective available potential energy (CAPE) is introduced in the final analysis, and a more intense supercell is predicted. A 50‐m tornado‐resolving simulation is conducted by nesting within the best‐performing 450‐m forecast. The simulation successfully captures the genesis and evolution of a tornado reaching EF2 intensity, through the one‐cell, two‐cell, and multivortex stages of development. The track, life cycle, and in‐storm location of the simulated tornado agree well with observations.