A numerical study on the impacts of lightning data assimilation upon the simulation of a double rainband in Fujian, China

Lightning data monitored by a very low frequency/low frequency (VLF/LF) three-dimensional lightning detection and position system were assimilated using a mesoscale numerical model and its three-dimensional variational assimilation system. A case study was conducted on a double rainband weather event in Fujian, China, on 7 May 2018, which included both shear-line precipitation and warm-sector rainfall. The impact of assimilating lightning data at cloud resolving scales was investigated through a series of comparative experiments, involving situations with and without lightning data assimilation, assimilation at different times, and assimilation under various parameter combinations. The results showed that assimilating lightning data significantly improved the simulation of the southeastward-moving shear-line rainband in inland Fujian and the northeastward-moving warm-sector rainband along the Fujian coast. The simulation of the warm-sector rainstorm was improved greatly. This improvement in simulating the coastal warm-sector rainstorm through lightning data assimilation was attributed not only to the increases in the relative humidity and cloud microphysical quantities, but also to the enhancement of wind convergence along the southwest jet in the ultra-low-level layer over the Fujian coastal region. The simulation of heavy rainfall in the double rainband was improved both in the individual assimilation of lightning data at 0200 and 0800 BT (Beijing Time), as well as in the simultaneous assimilation of lightning data at these two times. The modelling with assimilating lightning data at 0200 BT produced a greater improvement than that at 0800 BT. Moreover, the results from the simultaneous assimilation at both times were superior to those from individual assimilation at a single time. The simulation of the double rainband was also improved when assimilating lightning data under various parameter combinations, each yielding distinct improvements. The threat scores (TS) for 12-h rainfall were increased greatly, with the missing and empty report rates heavily reduced. The simulation accuracy for moderate rain, heavy rain, rainstorm and heavy rainstorm all improved obviously. Additionally, the TS for 1-h rainfall also increased, with a positive effect extending to 12–18 h. • The lightning data assimilation significantly improved the simulation of a double rainband weather event at the cloud-resolving scale. • The simulation of warm-sector extremely heavy rainfall, which is a challenge for worldwide operational numerical models, was significantly improved. • Assimilating lightning data two times sequentially gave a better simulation for the warm-sector extremely heavy rainfall than one time. • Sensitivity experiments were carried out on the parameters in the lightning data assimilation scheme to identify the best combination of parameters.