(BRAVO) Study: Final Report. - Desert Research Institute
(BRAVO) Study: Final Report. - Desert Research Institute
(BRAVO) Study: Final Report. - Desert Research Institute
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>Final</strong> <strong>Report</strong> — September 2004<br />
concentrations in the eastern portion of the domain. 3 On the other hand, at the twelve<br />
stations where the mean normalized bias represented a lesser portion (on average ~30%) of<br />
the mean normalized error (33-78%), eleven are located in the vicinity of Big Bend National<br />
Park and/or near the US-Mexico border, showing a tendency of the model to underestimate<br />
more often in these regions compared to other regions of the domain. 4<br />
Particularly severe mean normalized bias and error are apparent in Figure 9-25 at two<br />
locations in southern Texas, quite distant from Big Bend. Since poor model performance in<br />
southern Texas is not necessarily indicative of poor model performance at Big Bend, the<br />
impact of these two locations on the statistics in Table 9-19 was explored. Since Table 9-19<br />
is based on 37 stations, removing two stations would not be expected to have a large effect,<br />
and this turns out to be true. Specifically, for sulfate the effect is to slightly reduce the mean<br />
concentration to 3.91 µg/m 3 , increase r 2 to 0.49, and slightly decrease the bias and the error<br />
metrics. The biggest relative change is a decrease in mean normalized error from 65% to<br />
61%. The effects on PM 2.5 statistics are even smaller.<br />
For the duration of the <strong>BRAVO</strong> study and across the entire <strong>BRAVO</strong> Network,<br />
particulate sulfate accounted for 62% of the observed total sulfur concentrations on average,<br />
comparable to a sulfate fraction of 61% for the predicted total. About 3/4 of the predicted<br />
sulfate fractions lie within 20 percentage points of the observed values. Both data sets show<br />
that high sulfate concentrations tend to be associated with high sulfate fractions.<br />
This discussion indicates that model performance for sulfur differed in the eastern and<br />
western parts of the modeling domain. We can explore this difference by comparing<br />
performance at the K-Bar and Big Thicket stations. The K-Bar station at Big Bend National<br />
Park, apart from being the focal station of the study, is representative of the region where the<br />
model was more likely to understate concentrations of sulfur species. The Big Thicket<br />
station in eastern Texas is representative of the eastern portion of the domain, where the<br />
model showed high positive bias in simulated concentrations of sulfur species due to<br />
consistent overestimates in the region. Tables 9-20 and 9–21 provide performance statistics<br />
for these two locations.<br />
3 The predicted signal at the eighth station, Eagle Pass in southern Texas near the Mexican border, is<br />
affected detrimentally by inclusion of the Carbón power plants in the same model grid cell.<br />
4 At the twelfth station, Wichita Mountains, less than 30% of the potential data set was valid, which<br />
precluded any significant analysis.<br />
9-56