Impact of Radar Rainfall Data Assimilation on Short-range Quantitative Precipitation Forecasts using Four-dimensional Variational Analysis Technique

In Hong Kong, intense precipitation associated with the summer monsoon, monsoon troughs and tropical cyclones poses a major weather threat in the warm seasons. For quantitative precipitation forecasts (QPF) in the short range, the Operational Regional Spectral Model at the Hong Kong Observatory currently relies heavily on radar-based rainfall data ingested through a physical initialization process. Using the parallel (MPI) version of the same model, numerical simulations were conducted to explore the effectiveness of four-dimensional variational (4DVAR) data analysis technique in the assimilation of radar-based precipitation data. The impact on short range QPF was assessed by reviewing a collection of heavy rain cases in 2004 and 2005.

In the study of the selected cases, a more realistic analysis from the 4DVAR technique produced forecast fields that were able to depict the underlying forcing mechanisms for the observed intense convection. In the severe rainstorm case on 8 May 2004, explosive development in the vicinity of Hong Kong was shown to be triggered by mesoscale cyclogenesis at the low levels. In a more recent rainstorm case on 6 May 2005, the well-organized rainband passing over Hong Kong was shown to be the result of strong synoptic forcing associated with a trough of low pressure in a strong baroclinic environment. In the case of Severe Tropical Storm Kompasu in July 2004, vortex structure in terms of spiral rainband evolution and peripheral convective development was better predicted. After landfall, the vertical coherence of the vortex was retained for a longer period in the set of 4DVAR-based experiments, possibly explaining the observed rejuvenation of convection to the northwest of Hong Kong.

Despite the range of forcing mechanisms and rain intensity among the selected cases, the numerical simulations indicated that the 4DVAR analysis technique in general produced superior results in both the analysis and forecast precipitation fields, especially in the early hours of simulation. The associated mesoscale features were better organized and aligned with respect to the radar reflectivity and satellite images. Convection in terms of peak intensity and the relative positioning of rain areas were also better represented and forecast using the 4DVAR technique.

The study fully demonstrated the capability of 4DVAR technique in assimilating radar-based rainfall data effectively and, as a result, the promising outcome in short range QPF. Verification results comparing 4DVAR performance against physical initialization were compiled for reference. The relative merits of introducing satellite-based precipitation data in addition to or in place of radar-based rainfall using 4DVAR were also explored, and some future directions put forward for further discussion and investigation.