. 2012 Jun 25;46(14):7612–7620. doi: 10.1021/es3010305

Table 1. Model Performance Statistics for GSH OP and PM₁₀ Models.

pollu-tant	grouping	cross-validation R²^a					RMSPE^a ^b
GSH OP	across all	0.44					0.29
	by season	winter	spring		summer	autumn	winter	spring		summer	autumn
		0.25	0.34		0.39	0.34	0.30	0.28		0.29	0.32

	by year	2002	2003	2004	2005	2006	2002	2003	2004	2005	2006
		0.39	0.48	0.41	0.47	0.34	0.30	0.25	0.29	0.29	0.31

	by urban volu-metric density^c	low	medium		high		low	medium		high
		0.34	0.50		0.42		0.31	0.27		0.30

	by site type	urban background	roadside		kerbside		urban background	roadside		curbside
		0.45	0.36		0.13		0.25	0.32		0.31

	annual- mean^d	0.67					0.14
	spatial^e	0.73					0.13
	temporal^e	0.71					0.12

PM₁₀	across all	0.83					4.62
	by season	winter	spring		summer	autumn	winter	spring		summer	autumn
		0.85	0.86		0.81	0.73	4.24	5.15		4.12	4.56

	by year	2002	2003	2004	2005	2006	2002	2003	2004	2005	2006
		0.68	0.92	0.79	0.78	0.73	4.68	4.74	4.30	4.16	4.93

	by urban volu-metric density^c	low	medium		high		low	medium		high
		0.74	0.82		0.87		5.51	4.73		4.27

	by site type	urban background	suburban	roadside	curbside		urban background	suburban	roadside	curbside
		0.88	0.90	0.80	0.78		3.64	3.43	4.99	5.74

	annual- mean^d	0.67					2.68
	spatial^e	0.61					2.65
	temporal^e	0.76					2.16

Across weekly averages (2118 at 34 locations for GSH OP and 12 041 at 66 locations for PM₁₀). For PM₁₀, two high values at two sites were removed as outliers.

RMSPE is root mean squared prediction error. Units are OP μg^–1 for GSH OP and μg m^–3 for PM₁₀.

Defined using tertiles of the urban volumetric density (see text for definition) within 500 m. For PM₁₀, across only 6710 weekly averages at the 34 locations where GSH OP was also measured.

To calculate “annual-mean” values, measurements and predictions were averaged by year at each site with at least 31 weeks of data per year.

To calculate “spatial” values, measurements and predictions were averaged over time at each site with at least 40 weeks of data; similarly “temporal” refers to averaging over locations for all weeks for which measurements were available.

Table 1. Model Performance Statistics for GSH OP and PM10 Models.

Table 1. Model Performance Statistics for GSH OP and PM₁₀ Models.