Fig. 3
Performance of the PPDIST dataset, the ERA5 reanalysis, and the IMERG satellite product in estimating the daily P occurrence and peak P intensity indices. For the PPDIST dataset, we calculated the mean of the ten validation scores (one for each cross-validation iteration). The training scores were not shown as they were nearly identical to the validation scores. All scores were calculated using square-root-transformed observed and estimated values. Supplement Fig. S4 presents an equivalent figure for the 3-hourly P indices.
