Key Contributor to the High-impact Heavy Rainfall Event in Beijing China on 21 July 2012: TIGGE Ensemble and Targeted Observation Analyses

Key role behind the famous record-breaking heavy rainfall event that happened in Beijing China on 21 July 2012 was examined by ensemble-based correlation and targeted observation analyses. On 21 July 2012, more than 90% of Beijing metropolitan area was covered by 24-h rainfall of >100 mm with the mean 24-h rainfall of 190 mm over the whole metropolitan area and the maximum value of 460 mm. A total of 79 people were killed mostly due to the flooding and mud slide caused by the heavy rainfall.

The heavy rainfall was found to be a result of the interaction of multi-systems including an upper-level jet, a westerly mid-level trough, a low-level mesoscale convective vortex (MCV), subtropical high, low-level jet, and a landfalling tropical cyclone (Vicente). The southeast flow between the subtropical high and tropical cyclone Vicente and southwest low-level jet provided sufficient moisture toward Beijing area. It is interesting to know what the key contributors were and how they were related to the happening of this heavy rainfall event.

The key factor contributing to the heavy rainfall were first examined by calculating linear-correlation coefficient between 24-h area-average precipitation and various weather components using operational global ensemble forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF) provided by TIGGE. The results were further verified by comparing the good ensemble members with the bad ones identified using the Threat Score (TS) of the 24-h forecast precipitation. Results showed that the heavy rainfall was closely related to the strength and location of the MCV and the trough above it. Stronger MCV and deeper trough would produce heavier rainfall.

The impact of the MCV was further confirmed by targeted observation analyses using a strategy considering nonlinear error growth named “Conditional Nonlinear Optimal Perturbation” (CNOP). CNOP strategy produces a type of initial perturbation that has the largest nonlinear error growth in a verification area at the forecast time. This method can be used to identify the sensitive area where there are the maximum value of CNOP, or in other words where assimilating extra observation may improve the most forecast performance in a verification area at the forecast time. The sensitive area identified using CNOP moved with the location of the MCV and the trough above it with the changing of the leading time. Vertical distribution of norm (total energy) peaked in the low and middle level, which was consistent with the altitudes of the MCV and the trough.