For the first time since we started receiving this data, we finally had a chance with the November data set to examine the FTIR spectra on both the North and South fencelines. Although we had been provided spectra from the South FTIR for the previous 3 months, this data had largely been of very poor quality, and it was not until now that we have been able to take the time to thoroughly examine a full month's worth of data from the South FTIR monitor. The results of our our analysis have been very enlightening.
Spectra from both fencelines were analyzed for the following chemicals: MTBE, 2-methyl pentane, 3-methyl pentane, n-heptane, propane, and carbonyl sulfide. The results of this analysis indicate that there are some notable differences between measured emissions of these compounds on the opposing fencelines. With this November data set, the differences between the overall emissions "picture" on the North and South fencelines are becoming much clearer. . . clearer certainly than we get from the Tosco reports alone.
Our review of the FTIR spectral files for the month of November indicates that there is significantly more emissions activity along the South (supposedly upwind) fenceline than there is on the North. This is not something that anybody had predicted, and there is nothing in Tosco's monthly FTIR data summaries to suggest that this is the case. In some of the earlier reports, we have seen that the UV monitors were reporting more activity on the South than on the North, but the accuracy of this information has been repeatedly called into question, and in the absence of a QA program from the refinery, we still have no way to verify this data.
The North fenceline data is characterized by long periods of non-detect interspersed with infrequent hits of the target compounds in with average reported concentrations in the 100-200ppb range for most compounds. This data from the North is consistent with what we have seen in previous months, and continues a clear trend where the vast majority of reported hits occur in the early morning hours.
The FTIR data from the South fenceline took considerably more effort to sort out. This is partly because we had never worked with this data from this monitor before, (and therefore lacked an existing background reference file) and partly because the emissions picture along this fenceline is simply more complicated.
On this South fenceline, we detected the same target compounds much more frequently, but at much lower levels, than we have become used to seeing on the North. For example, we detect MTBE 862 times with a peak concentration of 59ppb on the South fenceline compared to just 162 hits with a peak of 287ppb on the North. There is also some evidence to suggest that MTBE is present along the South fenceline for long periods during the month at levels just below the detection limits of this equipment. We don't see this on the North.
Selecting a background reference spectrum for use on the South
As with any new data set, our first challenge with the November data was to identify a pool of "candidate spectra" from which we could select one spectral file for use as a background reference. This background selection process was made more difficult in this case because we were unable to identify any one background candidate spectrum which worked well for all of the compounds we were trying to measure.
The background file (Tos00727 from 11/13 at 10:01am) was found to be most effective for most of the compounds we were measuring, however, this background proved to be not as useful for quantifying one chemical, MTBE.
By comparing absorbance spectra which were created with Tos00727 with those created by using Tos00538 as a background reference, we can see that those files created with the 00727 background consistently report 10ppb less of MTBE than do the corresponding spectra made from the 00538 background. From this we can infer that the 00727 file contained approximately 10ppb of MTBE. More accurately stated, it appears that there was at least 10ppb of MTBE present along the South fenceline when this spectrum was generated.
The Tos00538 spectrum clearly represents a period of lower MTBE concentration, and does appear to serve as a better background reference for the purposes of detecting and quantifying MTBE concentrations. This spectrum, however, does not seem to work as well as a background reference for the other compounds we are monitoring (in the C-H Stretch region). It is not clear exactly why Tos00538 does not work well as a background reference in the C-H Stretch, as we have no direct indication that this spectrum contains any level of the various hydrocarbon molecules monitored there. It seems most likely, however, that Tos00727 simply possesses greater overall spectral quality, and provides for less noise and better detection limits in the C-H Stretch region . Visual comparison of corresponding absorbance spectra created by the two different background references (see fig. 1.) indicates that these two background files must be very similar indeed, as these absorbance spectra appear to be virtually identical.
For this data set, Tosco's contractor, Terra Air Services, reports a peak concentration of MTBE at 0.076ppm. That same data point (from 12:39am on 11/4) also represents the peak value of MTBE in our report (0.058ppm). By examining the databases we can see that the "uncertainty" or "3 Sigma" value which Terra reports for this data point is +/- .024ppm. That is, their reported concentration is 0.076ppm (+\- 0.024ppm). We report 0.058ppm (+\- 0.008ppm). Because our respective measurements fall within each other's "error bars", there is not necessarily a discrepancy here.
Because we do not know what Terra is using for a background reference, we can only guess as to the reason why they are measuring higher concentrations of MTBE here. It seems likely that they are using a background reference file which is "cleaner" than the one we were able to identify with respect to MTBE content. The relatively high uncertainty values and detection limits which they report here would seem to indicate either a problem with the overall quality of their background spectrum or with their analytical "script file". Either of these problems would help to explain why Terra was able to detect MTBE on the South fenceline only three times during this month.
In this data set we have identified 2-methyl pentane, 3-methyl pentane, n-heptane, and propane. All of these hydrocarbon compounds were identified via their absorbance features in the "C-H Stretch" region of the infrared spectrum centered near 2,900 wavenumbers. At times, all of these compounds are detected simultaneously. At other times, all of them are detected individually and in various combinations. Sorting out the various chemicals in this region has proven to be very challenging task, and there is a very real possibility that there are other chemicals (either that we have failed to identify correctly, or that are not available in our library of infrared reference spectra) which may be contributing to the concentration figures we report here. In contrast to the "Fingerprint Region" centered near 1,000 wavenumbers, the C-H Stretch is characterized by several chemicals which all display very similar absorbance features. A further explanation of this is available in the Sample Infrared Spectra page.
One of the biggest difficulties we face comes from the fact that these chemicals all can interfere with the computer's ability to identify and measure other chemicals with similar absorbance features. For example, if 2-methyl pentane appears in an absorbance spectrum - by itself - the computer may simultaneously identify this as 2-methyl pentane, 3-methyl pentane, n-heptane, n-hexane, etc., that is if 2-methyl pentane has not been identified as a potential interferent in the region(s) of analysis being used to measure these other compounds. (see figure 2.) Another example of the problem with interfering compounds is illustrated in (figure 3) where both 2-methyl pentane and 3-methyl pentane are identified as being present. Here, we can see that concentrations of both of these chemicals will be under-reported unless both of them are added as potential "interferents" to the other in the analytical "script matrix".
As we have acknowledged earlier, there is a very real possibility that we have made errors in our identification and quantification of these chemicals. We believe that with this November '97 data set, we are getting much closer to an accurate assessment of the chemicals which show themselves in the C-H Stretch. We still cannot say with any certainty, however, that there are not yet-to-be-identified chemicals which may be contributing to the concentration figures we are reporting here. What we can say for certain is that we see quite a bit of activity in this "C-H Stretch" region on Tosco's South fenceline in November '97. While we report some 341 "hits" of 2-methyl pentane, for example, it is important to point out that there are hundreds of other spectra in which no chemicals are reported - where we can see clear indications of chemicals absorbing light in this region (see fig. 4.)
A review of Tosco's report will give the reader no indication that there was any activity in this region whatsoever. We believe that our supplemental report, flawed as it may be, will provide a more complete characterization of the chemical emissions along Tosco's Southern border than can be ascertained from reading Tosco's report alone. We are making our various analytical "script files" available here to anyone who wishes to view them. These print-outs will likely be of little interest to anyone who is not a spectroscopist working on this project. We include them here in case Tosco's consultants choose to review our work, or in case there is anything in our script files which they could use in attempts to improve their own analytical methods. North Script . . South Script