Abstract
A short-term precipitation prediction (nowcast) method was developed by using upstream low-level humidification (ULH) in a cloud-resolving atmospheric model simulation. The ULH plays a role in the approximated adjoint calculation for time integration of four-dimensional variational data assimilation. This method was applied to a heavy rainfall event observed in the Kanto Plain on September 2, 2013. The heavy rainfall caused by a few cumulonimbus clouds was predicted by the ULH well before 30 minutes. The upstream advection period of 20 minutes was appropriate for more accurate prediction. The predicted cumulonimbus cloud included a large vertical vorticity that seems to have been associated with the observed tornado.
