Results of Model-to-Monitor Comparison for Benzene

Note: EPA no longer updates this information, but it may be useful as a reference or resource.

National-Scale Air Toxics Assessment
Results of Model-to-Monitor Comparison for Benzene
The model-to-monitor comparisons for benzene show good agreement between concentration estimates and monitored values. The model’s estimates are within a factor of two of the monitored concentrations for about 90% of the benzene monitors in this study. The scatter plot below shows the distribution of points between the 1:2 line and the 2:1 line.

For all sites which fall between the 1:2 line and the 2:1 line, the model was "within a factor of two": that is, the model-to-monitor ratio was between .5 and 2. This was true for 77 of the 87 benzene sites (89%). Even the model estimates that were not within a factor of two were close: all 87 sites had estimates within a factor of 3.

There are several reasons why we would expect the model and monitors to agree well for benzene:

There are a significant number of monitoring sites for it (105, for now), giving an adequate sample size for the statistics in the comparison.
Monitoring technology for it has a long history, suggesting that the monitoring data is reflective of actual ambient concentrations.
Its emissions have been tracked for a long time, so there is some confidence in emission estimates.
It is a widely distributed pollutant which is emitted from point, area, and mobile sources. Thus, if the model is biased in the way it handles any one of these source categories, the bias will be dampened.
We have an estimated background concentration for it.

Some of the statistics we will present in the final paper are below:

Pollutant	Number of Sites	Median of Ratios¹	Standard Deviation of Ratios²	Model Range Below Monitor (%)³	Model Range Includes Monitor (%)⁴	Model Range Above Monitor (%)⁵	Spearman's Correlation Coefficient	P-Value Associated With Correlation⁶
Benzene	87	0.92	0.42	36.78	43.68	14.54	0.67	.0000

Some of these variables are defined below:

(1),(2): the median and standard deviation of the 105 model-to-monitor benzene ratios.
(3),(4),(5): deal with the relationship of the range of census tract model estimates for the county (10th to 90th percentile) to the monitor average. (3) is the percent of sites for which the range is below the monitor average; (4) is the percent of sites for which the range includes the monitor average; and (5) is the percent of sites for which the range is above the monitor average.
(6): the P-value resulting from the statistical test that Spearman's correlation coefficient is greater than zero.

For more details on these statistics, please consult the model-to-monitor protocol paper.

In general, the statistics for benzene suggest good agreement between the model and the monitors. The median of the ratios is close to one, with a small standard deviation; this suggests that on average, the model estimates and monitor averages are close to each other. For almost half the sites, the range of county model estimates covers the monitor average; some of the remaining sites are missed on the low side and some on the high side, suggesting no clear bias. The last two variables show a clear positive correlation between the model estimates and the monitor averages.

Results of Model-to-Monitor Comparison for Benzene

Local Navigation