The Tokyo metropolitan area is characterized by a conspicuous heat island, but its influence on precipitation is still under controversy, as the majority of rainfall events are associated with large-scale disturbances. In this study, the long-term precipitation change at Tokyo was analyzed by digitizing hourly precipitation data from 1890 to 2007. At the same time, 30-year data (1978-2007) on the network of AMeDAS (Automated Meteorological Data Acquisition System), having a horizontal resolution of 15-20km, were used to capture the spatial anomaly of precipitation in Tokyo. Since the main target of the analysis was in localized, short-term showers that are most likely to be affected by urban processes, the analysis was focused on newly developing precipitation preceded by a period of no or little precipitation, hereafter denoted by "non-continued precipitation (NCP)". The analysis was made for six-hourly precipitation, obtained by a running sum of hourly values, so that NCP was defined by a case preceded by a six-hour period in which precipitation amount was less than 1mm.

The result indicates an increasing trend of NCP at a rate of 30%/century or more in late afternoon to early evening of the warm season. At the same time, NCP in Tokyo has a spatial anomaly of 30% or more, in comparison to the surrounding region, in the evening of the warm season. On the other hand, the trend and anomaly for all the precipitation cases are much smaller (~10%/century at most) than for NCP cases. These facts indicate that the increase of precipitation in Tokyo is characteristic to short-term precipitation in the afternoon of the warm season, in consistent with our understanding that thermal and dynamical forcing of the urban heat island is strongest in the daytime of summer.