JMA operates three types of objective analysis to make initial conditions for the operational numerical weather prediction (NWP) models. Those are Global Analysis (GA), Regional Analysis (RA) and Meso Analysis (MA). GA provides initial conditions for Global Spectral Model (GSM), RA for Regional Spectral Model (RSM) and MA for Meso Scale Model (MSM). The observational data assimilation system for GA is three dimensional variational and those for both RA and MA are four dimensional variational (4D-VAR). The operation of 4D-VAR for MA has been started in March 2002, while one for RA started in June 2003. The reason which MA preceded is that MSM is utilized for very short range rainfall forecast to mitigate natural disasters and it is essential to prepare highly balanced initial conditions to reduce the model spinning up time. The employment of 4D-VAR brought other various benefits. It made possible to make the most of hourly observations, time integrated observations and the observations not performed at analysis time. With the benefits, MA is assimilating hourly wind profiler data and one hour accumulated rainfall amounts. And the third benefit is much convenient for satellite observations. Satellite observations are essential to forecast rainfall in Japan since the islands are surrounded by ocean. There are few observations over there but the observations from satellite, although the system cause heavy rain crosses over there. Therefore, JMA had been developing the satellite data assimilation system for MA. TMI and SSM/I is polarized microwave imagers and there are a number of studies to retrieve the various meteorological values such as rain rate, precipitable water, sea surface wind speed, and so on. Some numerical weather prediction centers are developing the system to use the retrieved values or the brightness temperatures directly. However, it needs to solve radiative transfer models to assimilate brightness temperature and the computational cost is high. For the operational meso scale model, it is essential to finish the analysis in shorter time. From the reason, JMA select to use the retrieved rain rate and total column precipitable water in MA. To retrieve the values, the method developed by Takeuchi (1997) is used. The method is an empirical method to estimate those over the ocean and the computational cost is low. Observational system experiments (OSE) were performed for two weeks in June 2003. In the rainfall forecast, the threat score show positive effect of the data after 12 hours forecast, though the forecast before 12 hours are almost neutral. The reason is considered that the water vapor field over ocean is improved with the new data and it needs time to flow over Japan islands. Some case study showed good result forecasting heavy rain. With the result, JMA decided to use it in operational MA, and started it from 00UTC 15 Oct. 2003. The assimilation is running well right now.