Reliable rainfall prediction is a key for successful early warning of rain-induced long-traveling landslides. This study assesses the rainfall prediction skill of a high-resolution weather prediction model named the Multi-Scale Simulator for the Geoenvironment (MSSG). The rainfall accumulation is more crucial rather than hour-to-hour rainfall intensity for long-traveling landslide. Three heavy rain events, which caused devastating landslide events in Sri Lanka, are chosen for the assessment: (i) Aranayaka in 2016, (ii) Rilpola in 2017 and (iii) Kiribathgala in 2014. The raingaunge observatory data are compared with the 2 km-resolution MSSG simulation and the Global Satellite Precipitation Map (GSMaP) data with approximately 10 km resolution. Comparison results show that the high-resolution MSSG simulation can represent the 3 day and 24 h rainfall accumulations and also the maximum rainfall intensities better than the GSMaP data. In addition to the assessment for rather short-term prediction, we investigate the exceedance probabilities (EPs) of daily maximum of hourly rainfall intensities at two adjacent raingauge points. The EPs show large difference between the mountainous point and plain point even though their separation is just 4.6 km. The 2 km-resolution MSSG simulation can capture the difference in the two EPs. This confirms the MSSG’s reliability in capturing the topographic enhancement of rainfall intensities.