Skip to main content
. Author manuscript; available in PMC: 2021 Feb 1.
Published in final edited form as: Atmos Environ (1994). 2019 Nov 14;222:117130. doi: 10.1016/j.atmosenv.2019.117130

Table 5. Overall performance of exposure assessment methods by number of nearby stations.

Correlation coefficient between PM2.5 concentration predictions and observed PM2.5 concentration in μg/m3 stratified by the number of active monitoring sites within 50 miles of the prediction site for each day. The methods considered are: raw CMAQ output, ordinary least squares (“OLS”), inverse distance weighting (“IDW”), universal Kriging(“UK”), downscaler, random forests (“RF”), support vector regression (“SVR”) and Neural networks (“NN”). Methods use either CMAQ and/or other geographic covariates (“Covs”).

Active nearby stations <5 5–9 10–19 ≥ 20
CMAQ 0.50 0.56 0.59 0.59
OLS (CMAQ) 0.60 0.69 0.73 0.77
OLS (Covs) 0.70 0.80 0.81 0.85
OLS (CMAQ+Covs) 0.64 0.74 0.75 0.79
IDW 0.82 0.91 0.92 0.90
UK (CMAQ) 0.84 0.92 0.92 0.92
UK (Covs) 0.83 0.91 0.92 0.92
UK (CMAQ+Covs) 0.81 0.91 0.91 0.91
Downscaler(CMAQ) 0.83 0.92 0.93 0.92
RF (CMAQ + Covs) 0.69 0.78 0.81 0.84
SVM (CMAQ + Covs) 0.74 0.86 0.87 0.87
NN (CMAQ + covs) 0.74 0.83 0.85 0.86