An intercomparison of three satellite infrared rainfall techniques over Japan and surrounding waters

Three satellite infrared techniques of increasing complexity for the estimation of rainfall are compared: (1) the GOES precipitation index (Arkin, 1979, 1983), (2) the Negri-Adler-Wetzel (Negri et al., 1984) technique, and (3) the convective-stratiform technique (CST), described by Adler and Negri (1988). It is shown that all three techniques performed poorly in estimating the rain maxima over southeastern Japan associated with shallow orographic warm rain systems. However, when the data set was limited to the ocean-only points, the CST method performed best, with the lowest bias, the lowest rms, and the highest correlation. Daily rainfall estimates had rms errors of almost 200 percent of the mean, and negative biases of about 50 percent of the mean.