Numerical Weather Prediction (NWP) by the Japan Meteorological Agency (JMA)
The Japan Meteorological Agency began numerical weather prediction in June 1959. Since then, NWP model performance has advanced significantly thanks to progress in earth sciences and information technology (e.g., dramatically improved computer resources and efficient telecommunication systems) as well as improved observation systems (especially those involving the use of meteorological and earth-observing satellites).
JMA currently operates several NWP models to cover various types of prediction, including very-short-range forecasts, short-/medium-range forecasts, typhoon track forecasts and aviation forecasts. The Agency also operates ensemble prediction systems (EPSs) and a coupled ocean-atmosphere model for other forecasts, including one-week/one-month/seasonal predictions and El-Niño forecasts.
JMA works continually to improve its NWP systems. In March 2006, a Meso-Scale Model with higher horizontal resolution was introduced and the frequency of its operation was increased. In November 2007, the resolution of the global model was enhanced, and one-week EPS based on global model products was also improved. In February 2008, a new EPS targeting typhoon forecasts was launched.
In August 2012, the Local Forecast Model covering the eastern part of Japan was put into operation. In May 2013, its forecast domain was expanded to cover the whole of Japan and its operation frequency was raised from every three hours to every hour. In March 2013, the forecast domain of the Meso-Scale Model was also enlarged, and the range of the forecast runs made by the Global Spectral Model at 12UTC and the One-week Ensemble Prediction System was extended from 216 to 264 hours. In May 2013, the Meso-Scale Model's forecast period was extended to 39 hours for all eight daily operations (once every three hours).
In February 2014, JMA improved the horizontal resolution of the One-week Ensemble Prediction System from 55 to 40 km and increased the operation frequency and the number of ensemble members per day. In March 2014, it also improved the Global Spectral Model by increasing the number of vertical layers from 60 to 100, raising its top level from 0.1 to 0.01 hPa and revising the physical processes involved. In addition, the Typhoon Ensemble Prediction System was improved by increasing the horizontal resolution from 55 to 40 km and the number of ensemble members from 11 to 25.
Data assimilation systems for NWP are based on the variational method. The four-dimensional variational (4D-Var) systems was introduced for Meso-Scale Analysis and Global Analysis in March 2002 and February 2005, respectively. The Local Forecast Model is initialized with an hourly assimilation-forecast cycle running three-dimensional variational (3D-Var) analysis and one-hour forecasts in turn for three hours.