Table 3.

Model performance (cross-validated R² and RMSE) for PM_10–2.5 mass (μg/m³) and species concentrations (ng/m³)^a using land use regression (LUR) and universal kriging (UK) models.

Model	CV measure	Total mass	Copper	Phosphorus	Silicon	Zinc
aSilicon values in μg/m3. bThe predictors included in each model are listed in Supplemental Material, Tables S2–S6. bFor pooled models across all cities, we present the R2 for the explanatory power of each city as calculated by the formula (1 – sum of squared differences between observations and predictions/sum of squared differences between observations and city-specific means). R2 with values < 0 are reported as 0. RMSE was calculated as the square root of the average of squared differences between observations and predictions. The respective R2 and RMSE values for each city appear in the following order: Chicago, St. Paul, Winston-Salem.
Land use regressionb
Chicago, IL	R2	0.68	0.65	0.50	0.68	0.73
Chicago, IL	RMSE	1.16	2.29	3.88	0.08	10.63
St Paul, MN	R2	0.51	0.86	0.68	0.93	0.40
St Paul, MN	RMSE	2.33	0.61	4.14	0.07	4.44
Winston Salem, NC	R2	0.41	0.51	0.76	0.48	0.36
Winston Salem, NC	RMSE	1.09	0.93	3.95	0.07	1.89
All citiesb	R2	0.52, 0.54, 0.10	0.65, 0.49, 0.09	0.34, 0.59, 0.66	0.24, 0.64, 0	0.61, 0, 0
All citiesb	RMSE	1.39, 2.24, 1.33	2.26, 1.06, 1.25	4.39, 4.63, 4.66	0.13, 0.16, 0.12	12.72, 6.10, 3.58
Universal kriging
Chicago, IL	R2	0.68	0.64	0.50	0.68	0.73
Chicago, IL	RMSE	1.14	2.32	3.88	0.08	10.60
St Paul, MN	R2	0.51	0.86	0.68	0.91	0.38
St Paul, MN	RMSE	2.32	0.61	4.14	0.08	4.52
Winston Salem, NC	R2	0.41	0.51	0.76	0.47	0.36
Winston Salem, NC	RMSE	1.09	0.93	3.95	0.07	1.89
All cities	R2	0.51, 0.52, 0.11	0.20, 0.20, 0	0.15, 0.38, 0.65	0.22, 0.64, 0	0, 0, 0
All cities	RMSE	1.42, 2.29, 1.33	3.42, 1.34, 2.10	5.00, 5.67, 4.78	0.13, 0.16, 0.13	22.01, 7.66, 10.58