Abstract The topographical conditions of Sakurajima active volcano change due to accumulation of volcanic ash and erosion due to surficial flows. The accumulated water coming through precipitation and steep slopes of gullies initiates lahar carrying volcanic material which potentially causes serious problems in the near vicinity. The permeability of the ground surface changes within time due to accumulation and erosion of volcanic ash. In this study, rainfall intensity-duration (ID) curves which can be used as an early warning system were produced by considering the monthly ash fall thicknesses, lahar type, minimum inter-event time (MIT) and in terms of active crater years. In this regard, 62 lahars were investigated in the Arimura river basin of Sakurajima volcano in between 2015 and 2020. The rainfall data is gathered by X Band Multi-Parameter (XMP) radars to assess the precipitation in the catchment basin. Lahars are investigated by adopting kinematic wave method and categorized based on their waveforms. According to performance statistics by considering confusion matrices, the best possible ID thresholds are determined for different conditions. It is found that higher performance metrics are obtained when lahars are categorized by kinematic wave model. It is revealed that monthly ash fall thickness of 0.1 cm and above is the critical value when the ID thresholds dramatically decrease above this thickness. The best ID models result that the missed lahars are 9% and false alarms are 26%. Majority of the false alarms are observed only below 0.1 cm of monthly ash fall thickness indicating that the lahars occurred above this level can be estimated successfully by ID curves.