Dynamical and Numerical Studies of Mesoscale Meteorological Phenomena

Chief Scientist M. YOSHIZAKI

Various atmospheric phenomena can be simulated with a super-computer. As for mesoscale disturbances which are not easily understood by observation alone, we can make clear their developing and decaying mechanisms by changing various conditions in the simulation models. These experiments may improve our understanding of the structure and mechanism of mesoscale systems in connection with environmental state and/or orography. These results can be applied to the prediction of heavy rainfall and heavy snowfall.

We are also developing data assimilation methods (physical initialization, diabatic NNMI etc.) to incorporate remotely-sensed data (measured by DMSP SSM/I, NOAA AVHRR, GMS5 etc.) into mesoscale numerical weather prediction models.

Main themes of the studies are:

MRI-NHM Home Page
MCS Home Page



Fig. 1 Observed precipitation intensity field, calibrated by radar at 1245 JST, 1330 JST and 1415 JST. Fig. 2 (a) Observed precipitation intensity field, calibrated by radar with a 7.5-minute interval. (b) Predicted vertical velocity fields with a 10-minute interval at z* = 1.30 km. The predicted q field near the surface is also shown by solid contours. The arrows represent the movements of convective cells.


Torrential heavy rain occurred over southern Kyushu on 1 August 1993. Recored rainfall was over 30 mm hour-1 which fell continuously from 12 JST (JST = UTC + 9 hours) to midnight. During the first stage of this heavy rain, a band-shaped stationary rainfall area was observed over the Baiu frontal zone.

The observation of the radar showed that many convective rolls parallel to the east-west direction lay scattered widely at first, and then convective cells grouped into several systems. A band-shaped rainfall area parallel to the WNW-ESE direction was produced for two hours (Fig. 1). The rainband was consisted of meso-scale convectives systems with the line-shaped structure, whose orientation was different from that of the rainband. The movement ofconvective cells showed that individual cells moved almost westward. This direction corresponds with low-level wind direction. Developed convective cells moved with a speed of 15 m/s, while newly generated cells moved slowly. The repeated generation of convective cells has the characteristic features of the back-building type (Fig. 2a). The MRI-NHM with a horizontal grid size of 2 km successfully reproduced this back-building system (Fig. 2b). 2 km successfully reproduced this back-building system (Fig. 2b).

Back to Before Page

Development of Variational Data Assimilation Systems for Mesoscale Numerical Models

Chief Scientist K. SAITOH

Mesoscale numerical models with a horizontal resolution of a few to 10 km are being developed in the world for research and operational purposes. Those models can be used to predict mesoscale high-impact weather phenomena such as heavy rainfall, heavy snowfall and localized strong wind. The process of preparing initial conditions for numerical models by effectively utilizing observational data is called gdata assimilationh. The Forecast Research Department is developing data assimilation systems for mesoscale numerical models based on a variational principle. Those systems enable us to directly assimilate various observational data from surface observations, radiosondes, wind profilers, Doppler radars, GPS observations, satellite microwave observations and so on. The systems provide high-quality initial conditions and improve the forecast skill of mesoscale numerical prediction. The main theme of these studies are:

Impacts of Doppler radar radial wind data and GPS precipitable water data on a rainfall prediction for the localized heavy rainfall in the Kanto area on 21 July 1999, obtained from a 4-dimensional variational data assimilation system and a mesoscale numerical model with a horizontal resolution of 10 km. Left: observed precipitation rate at 18 JST. Middle: predicted one-hour precipitation during 17-18 JST and low-level wind at 18 JST when only the conventional data were assimilated (initial time: 15 JST). Right: same as the middle panel but when the radial wind and precipitable water data were also assimilated. The direct assimilation of those two additional data improves the forecast of rainfall area.

Back to Before Page

Studies on Objective Weather Forecast Techniques

Chief Scientist F. FUJIBE

This project covers studies for the development and improvement of weather forecast techniques to make forecast more objective and quantitative.

Main themes of the studies are:

This system classifies weather maps and predicts bad weather appearance, areas of bad weather and precipitation occurrence.

Back to Before Page