Table S2.

Error analysis of the gap filling of the methane fluxes

Values	BES	BEO	CMDL	ATQ	IVO
Number data points, n	8,350	6,978	6,891	6,930	6,309
Data coverage, %	60.0	50.0	49.5	49.8	45.4
Neurons	8	8	8	8	8
Runs	100/500	100/500	100/500	100/500	100/500
Final averaged runs	25	25	25	25	25
Daytime training data coverage, %	49	51	50	49	51
Nighttime training data coverage, %	50	48	49	50	48
Daytime testing data coverage, %	49	50	50	51	49
Nighttime testing data coverage, %	50	49	49	48	50
R2-training	0.91	0.88	0.61	0.76	0.57
R2-testing	0.89	0.83	0.52	0.75	0.45
Mean original, true units	0.56	0.41	0.26	0.25	0.89
Mean modeled, true units	0.51	0.43	0.27	0.25	0.86
Mean gap-filled, true units	0.51	0.43	0.27	0.25	0.86
Mean RMSE, true units	0.171	0.15	0.181	0.125	0.69

Error analysis results of the gap filling of the CH₄ fluxes for each of the indicated sites, including the coefficient of determination (R²) and the RMSE converted to true physical units of mg C-CH₄ m⁻² h⁻¹ indicating the uncertainties of the models.