Fig. 4: Statistical evaluations of the test set.
From: Machine learning-based tsunami inundation prediction derived from offshore observations
a Aida’s numbers (K and κ) and accuracy measure in percentage using maximum and mean tsunami amplitudes as inputs at three prediction time windows of 10, 15, and 20 min. b Distribution of misfits with 15th, 50th, and 85th percentiles (dashed lines). c Box plot of the goodness-of-fit (G) for all samples at each specified earthquake magnitude. b and c are based on the predictions using the 20-min window of maximum tsunami amplitude.
