Table 3.

Statistical analysis of modeled O3 concentration using surface, ozonesonde, aircraft, and satellite observations.

	N	Mean		R	NMB	NME
		Observation	Model
Surface
– WDCGG	1498	53.9	43.5	0.49***	−19.3%	23.7%
– CASTNET	2316	53.4	52.9	0.61***	−0.9%	12.6%
– EANET	240	56.2	49.1	0.49***	−12.6%	20.6%
Ozonesonde
US and Canada
– boundary layer	1016	46.0	42.7	0.70***	−7.1%	16.7%
– free troposphere	893	87.7	59.6	0.79***	−32.1%	33.8%
– upper model layer	512	905.1	770.3	0.91***	−14.9%	30.2%
Asia
– boundary layer	283	44.2	47.4	0.44***	7.1%	24.5%
– free troposphere	207	70.7	59.0	0.43***	−16.5%	21.5%
– upper model layer	124	529.8	399.2	0.94***	−24.6%	34.4%
Europe
– boundary layer	1478	47.6	46.8	0.42***	−1.6%	17.9%
– free troposphere	1368	78.2	57.2	0.76***	−26.8%	29.1%
– upper model layer	817	1015.7	894.3	0.94***	−11.9%	24.2%
Aircraft
– from the surface up to 6km	128	55.9	45.3	0.74***	−19.0%	19.1%
Satellite
– tropospheric column	28020	33.2	34.7	0.65***	4.7%	13.5%

Note: the unit of mean for observations and simulations is ppbv except satellite observation, which is expressed as DU. Maximum daily 8h average ozone (MD8O3) is used for surface observational data of WDCGG, CASTNET, and EANET. Corresponding hourly modeled O₃ is used for ozonesonde data. The 2–4h averaged hourly modeled O₃ is used for aircraft data to fully cover each observation time. Significance levels by Student’s t test for correlation coefficients between observations and simulations are _{marked as}

^*p<0.05_,

^**p<0.01_{, and}

^***p<0.001, and lack of a mark indicates no significance.