Table 3 Evaluation of the classification skills of our strategy against GPM, ERA5, its 2-km dynamical downscaled reanalysis VHR, and LR for the L1 region.

Metric	Task	1 h					12 h					24 h
		AI	GPM	ERA5	VHR	LR	AI	GPM	ERA5	VHR	LR	AI	GPM	ERA5	VHR	LR
Precision	Calibration	0.64	0.43	0.31	0.52	0.61	0.78	0.58	0.46	0.68	0.75	0.80	0.64	0.55	0.74	0.78
Precision	Regionalization	0.64	0.42	0.30	0.52	0.60	0.77	0.56	0.47	0.68	0.73	0.80	0.62	0.55	0.74	0.77
Recall	Calibration	0.62	0.50	0.81	0.41	0.59	0.72	0.64	0.86	0.56	0.70	0.76	0.71	0.89	0.62	0.76
Recall	Regionalization	0.64	0.53	0.81	0.42	0.61	0.74	0.67	0.87	0.56	0.72	0.78	0.74	0.89	0.62	0.76
F1	Calibration	0.63	0.46	0.44	0.46	0.60	0.75	0.61	0.60	0.61	0.72	0.79	0.67	0.68	0.67	0.78
F1	Regionalization	0.64	0.47	0.44	0.46	0.60	0.75	0.61	0.61	0.61	0.72	0.80	0.67	0.68	0.67	0.77
Accuracy	Calibration	0.93	0.88	0.80	0.90	0.92	0.88	0.79	0.71	0.81	0.86	0.85	0.76	0.70	0.78	0.84
Accuracy	Regionalization	0.93	0.90	0.79	0.90	0.92	0.88	0.78	0.71	0.81	0.86	0.85	0.74	0.70	0.78	0.84

L1-fold-averaged Precision, Recall, F1 score, and Accuracy are displayed for all the analyzed cases and models. Best results are highlighted in bold

Quick links

Search