Many radiative transfer models and physical rainfall retrieval schemes have an assumption that rainfall rates are homogeneous across the satellite field of view (FOV). Then, in most cases, we can see significant differences between precipitations derived by the plane-parallel radiative transfer model and observations due to this assumption. It has been called as the "beamfilling" effect. One of the most basic problems is that precipitations are not uniformly distributed within each pixel.

In this paper, the horizontal inhomogeneity of precipitations from radar data is investigated. Analyses of one-year-long AMeDAS (Automatic Meteorological Data Acquisition System) precipitation data in Japan and four-month-long ship borne radar data during TOGA COARE (Tropical Ocean Global Atmosphere Coupled Ocean-Atmosphere Response Experiment) in the western Pacific Ocean are performed. A great number of pixel-averaged precipitation rates corresponding to the satellite FOV is obtained over the whole domain and entire period of these data sets. According to the averaged precipitation rates, we can characterize the horizontal distribution from the heavy precipitation values to the light ones. The statistical characteristics of precipitations are described and the precipitation ratio (1-the number of no precipitation points/the total number of points) within the sub-pixel is also discussed.

By analyzing the characteristics of precipitation inhomogeneity, we can find the way to more accurately deal with the beamfilling effect by the inhomogenity problem of precipitations and this study may provide the important basis in the developments of radiative transfer models and retrieval schemes.