The objective of this study is to explore the applicability of very short-range-forecast rainfall for the early warning of
mud-debris flows. An artificial neural network was applied to use the very short-range-forecast rainfall data. The neural
network is learned by using the relationship between the radar and the AWS, and forecasted rainfall is estimated by
replacing the radar rainfall with the MAPLE data as the very short-range-forecast rainfall data. The applicability of
forecasted rainfall by the MAPLE was compared with the AWS rainfall at the test-bed using the rainfall criteria for
cumulative rainfall of 6hr, 12hr, and 24hr respectively. As a result, it was confirmed that forecasted rainfall using the
MAPLE can be issued prior to the AWS warning.