Highly time resolved fine particle nitrate measurements at the Baltimore Supersite

Nitrate in particles smaller than 2.5 mum was measured at 10-min intervals at the Baltimore Supersite in east Baltimore from 14 February through 30 November 2002, using the R&P 8400N semicontinuous monitor to determine its contributions to fine-particle aerosol mass concentrations. Comparison with 24-h filter-based measurements, revealed a discrepancy of 33% between the 24-h averages derived from the two methods, for most of the 9.5 month study period, despite corrections for conversion efficiency and Reaction Cell Pressure deviations, suggesting a true conversion efficiency of 68%. Estimates of precision in individual 10-min measurements averaged 8.7% and ranged from 6.3% to 23%, excluding uncertainty encompassing dissociation losses. Uncertainties in 24-h averages of the 10-min measurements were generally larger (median of 9.1%) owing to missing or invalid values. The detection limits for 24-h averaged and 10-min concentrations were typically 0.17 and 0.24 mug m(-3), respectively, during the study (both after slope correction to achieve agreement with 24-h speciation measurements). Regression slopes were statistically equivalent for all months except February and October (an outlier not understood). Intercepts were generally small and insignificant. Good agreement between the 24-h data sets was achieved after the monthly mean regression slopes were applied to the 10-min data. In February, when flat flash strips were used and instrument compartment/outdoor ambient temperature differences were often severe, the regression slope was statistically larger than the average for the remaining months and the intercept was positive and significant. Results of a nonlinear least squares model used to estimate dissociation losses suggest that the largest errors occurred when concentrations are near the detection limit, instrument-outdoor temperature differences were large, and ambient RH low (<40%), i.e., conditions which most frequently and severely occurred in February and March. In February, dissociation losses as large as 1.65 mug m(-3) (100% of the slope-corrected measured value) may have occurred and such losses were predicted to be > 30% in 63.5% of the 10-min measurements for that month. However, model predictions for the other months, when new ridged-flash strips were used, suggest that dissociation losses were much less significant, i.e., <15% in >95% of the measurements. Our experience suggests that the semicontinuous monitor can produce reliable 24-average concentrations when instrument-outdoor differences are kept small, an independent measurement is used to correct the data, and are improved when grooved flash strips are used. (C) 2004 Elsevier Ltd. All rights reserved.