During the intensive observation period (IOP) of the GAME, enhanced upper air and surface pressure observations are assimilated to improve the 4DDA products around the monsoon Asia. Enhanced surface observations including the Asian Automated Weather Station Network (AAN) are also effective for validating the land-atmosphere interactions expressed in the 4DDA products.

Continuous efforts are being made to develop data assimilation techniques of satellite data. Now, the ERS-1 and the DMSP are in operation of providing the surface wind over the ocean and total precipitable water, respectively. In a couple of years, new earth observing satellites ADEOS and TRMM will provide observation data of the surface wind over the ocean and precipitation, respectively. Assimilating these data must be an important research subject in several years.

The 4DDA products are open for all scientists participating in the GAME. The 4DDA products have physical consistency among variables and uniform resolution

in space and time so that they are convenient for making research on the mechanism of the Asian summer monsoon. Users are recommended to notice quality of the 4DDA products, because they tend to be affected by climate drifts of the NWP model. Intensive validation with special observations must be a great help to check quality of 4DDA products.

The 4DDA products and special observations are also used for verification of the long-range forecasts of the Asian summer monsoon, paying particular attention to cloud-radiation interactions, deep cumulus convections, land-atmosphere inter-actions and ocean-atmosphere interactions. Based on both of understandings of the Asian summer monsoon and intensive verifications by special observations, we will improve the long-range NWP models.

Objective analyses are performed on the same h-level and Gaussian grids as used in the global model with an optimum interpolation (OI) method, where the first guess field is 6-hour forecast of the global model running from the previous analysis (6 hourly forecast-analysis cycle). Geopotential height and wind are analyzed with a multivariate OI assuming the geostrophic balance, while temperature and moisture fields are analyzed with a univariate OI except that moisture is set equal to the first guess field above 300 hPa (forecast-forecast cycle).

Quality of the 4DDA products, to a great extent, depends on data quality control (QC) system and OI parameters. Thus, we systematically survey errors of all observational data and 6-hour forecasts (the first guesses to objective analyses) and update the QC and OI parameters.

Variational methods of objective analysis are expected to reasonably consider dynamical constraints of the atmosphere and solve inversion problems from observation data into the prognostic variables. Now, we are developing the three dimensional variational method (3Dvar) which can includes constraint of equilibrium equations such as gradient wind balance and assimilate the surface wind

observed by ERS-1 and ADEOS. After completing the 3Dvar in a couple of years, we will develop the 4Dvar in another several years. The 4Dvar enables us to include prognostic equations as constraints and assimilate precipitation data.

The observations of upper air and mean sea-level (or surface) pressure enhanced in time and space can be directly assimilated and they may be effective especially for diurnal variations and meso-scale disturbances in the monsoon Asia. Surface wind observed by ERS1 and ADEOS and total precipitable water by DMSP will be assimilated with the 3Dvar. The precipitation data from TRMM can be assimilated with the 4Dvar. The precipitation data might have large impacts on the analyses of moisture and horizontal divergence. The TRMM data, however, has some ambiguity in its retrieval and needs appropriate truth data.

In the GAME, valuable truth data are available from the AAN and enhanced surface stations. Observed data of the soilwetness, snow mass and soil temperature can be used in the data assimilation of the ground hydrological processes. In this regard, we must study the way of estimating mean values within the grid box of prognostic models from the observations, because they tend to represent local characters.

Model-generated precipitations and fluxes are important products of 4DDA in promoting the monsoon research. Quality of these products is significantly dependent on the performance of the NWP model. Thus, in order to improve the 4DDA products, we must refine the NWP model through validation with special observations. In this direction, the following projects are under planning;

(1) Surface water and energy cycles

Advanced observations of precipitation and surface fluxes including Bowen ratio can be used as truth data for validations of the 4DDA products. Fig.4.3-1 shows diurnal variations of precipitations, horizontal moisture flux convergence and evaporation obtained through the 4DDA. Although those diurnal variations look reasonable, they may be subject to climate drifts of the NWP model. The AAN and the TRMM satellite enable us to validate detailed diurnal variation and geographical distributions of precipitation, respectively.

(2) Radiation budgets

(3) Stand-alone time-integration of simple biosphere (Sib) model

In the JMA's global model, the ground hydrological processes are described by the simplified biosphere model. To advance the Sib model, we are making experiments on stand-alone long-term integrations of the model. Driving forces of the model are observed data of downward radiation, precipitation and surface wind, while validation data of the prognostic variables are those of soil moisture, snow mass and surface fluxes of latent and sensible heats. In this experiment, we can make maximum use of the AAN data focusing on land-atmosphere interactions in four different regions ranging from the tropics to the tundra.

(4) Long-range forecasts of the Asian summer monsoon

One of the ways of improving the 4DDA system is to reduce systematic errors of the NWP model used in the data assimilation. This meets just with improvement of the long-range forecast model. In March, 1996, JMA started one month ensemble forecasts by using a T63L30 version of the global model with the same physics parameterization schemes. Detailed analyses will be made of forecast performance of the Asian summer monsoon by comparing to the 4DDA products and individual observations. According to this, we will improve the parameterization schemes of the NWP model. For this purpose, we will also benefit from the understanding of the Asian monsoon achieved in the GAME.

Prognostic variables of the atmosphere

(2D: surface pressure, 3D: temperature, wind and humidity)

Radiative fluxes

Atmospheric water budgets

(2D: total precipitable water, total horizontal moisture flux)

Surface fluxes

(2D: latent and sensible heat fluxes, surface stresses, precipitation etc.)

Ground hydrological parameters

(2D: run off, soil water, snow, deep soil temperature, roughness etc.)

Diabatic forcings

(3D: heating rate due to radiation, PBL and condensation, moistening rate

due to PBL and condensation, acceleration due to PBL and gravity wave

drag)