AbstractWater vapour advection from the sea causes extremely heavy rainfall in Japan. Therefore, accurately describing the water vapour distribution over the sea in a forecast model's initial conditions should improve the prediction accuracy of heavy rainfall events. Thus, we assimilated the shipborne precipitable water vapour (PWV) observed by the Global Navigation Satellite Systems (GNSS) onboard ships to show its impact on a heavy rainfall event in July 2020. We obtained the GNSS observations during a continuous water vapour observation campaign conducted at sea, one of the few in the world. In this study, we applied the shipborne PWV, on the upwind side of the heavy rainfall, to the four‐dimensional variational data assimilation method and conducted assimilation–forecast cycles. Although the shipborne PWV is a point observation, it can be assimilated as observation data covering space and time because the ships conducting the GNSS observation sailed around Japan. In addition, in the assimilation–forecast cycle, the effect of assimilation spread widely over the forecast area. Although the ship motion affects the shipborne observations, we found that the impact on PWV assimilation is negligible for practical use. For the temporal thinning of the data assimilation slot for the shipborne PWV, an interval of 30 min is more effective than an interval of 1 hr. As a result of assimilating this shipborne PWV on the upwind side of the disastrous heavy rainfall of July 2020, the forecast accuracy of rainfall, especially rainfall amount, was improved. We also found that statistical improvements could be obtained from the water vapour profiles and wind velocity field in the lower atmosphere. We demonstrate that the assimilation of shipborne PWV observations can improve the prediction accuracy of heavy rainfall events.