This extreme precipitation event and its synoptic-dynamic causes are analyzed using satellite-derived rainfall estimates (NASA GPM IMERG), geostationary satellite imagery (JMA Himawari-8), radiosonde measurements, and the new ECMWF fifth generation reanalysis ERA5. The analysis suggests that the event was mainly caused by two factors:
- After breakdown of the daytime sea-breeze circulation over the Malay Peninsula on 25 September 2016, an associated convergence line moved eastward with the monsoonal westerlies and passed the Gulf of Thailand during nighttime. The convergence line reached the coast of southwestern Vietnam on the morning of 26 September and initiated convection over the Ca Mau peninsula in southern Vietnam as it continued to move northeastward towards HCMC.
- In association with Typhoon Megi northeast of Luzon (Philippines), northeasterly low-level winds occurred over the northern and central South China Sea on 26 September 2016, which were forced to flow around the southern parts of the central Vietnamese Truong Son mountain range. This helped to induce convection due to orographic lifting and due to confluence of northwesterly and northeasterly winds of the around-mountain flow with monsoonal westerlies.
A cold pool that was associated with convection near the southern parts of the Truong Son mountain range (2.) then collided with the convergence line (1.) in the vicinity of HCMC, leading to an intensification of convection and keeping convection rather stationary over HCMC. The stationarity could also be ascribed to the location of HCMC at the western end of the monsoon trough, thus a low-level steering flow was absent. The monsoon trough was pushed to an unusual equatorward position over the South China Sea due to the northeasterly flow in association with Typhoon Megi. While such a synoptic situation and the convergence line described under (1.) are rather regular features, the timing and location of convection over the Truong Son mountains, the generation of a mesoscale cold pool and its convergence with the convergence line to the east and over the city contributed to the rarity of the event. Likely due to the latter role of convection, neither the TIGGE multi-model ensemble forecasts, nor numerous convection-permitting simulations with the WRF (Weather Research and Forecasting) model were able to forecast the location and intensity of this extreme event. This demonstrates the need to develop nowcasting techniques that use extrapolation of radar data and the assimilation of the latter into convection-resolving nowcasting models.