· We detect propane at concentrations approaching 2ppm on the North FTIR on October 5, these and other detected concentrations of propane go completely unreported by Tosco’s contractors.
· We see continued reporting of higher than expected levels of several compounds, including carbon disulfide, by one of the UV monitors on the South.
· We see reports of hydrogen sulfide in parts per million on one of the North TDLAS units. Terra reports that these hits were determined to be "false positives". By examining the databases, it is unclear how they could have made such determinations.
· The South FTIR unit continues to provide very poor quality spectra, a problem which has been allowed to continue since mid-August.
The primary problem, from our view, with the fenceline system continues to be the complete lack of any type of stated Data Quality Assurance program or schedule on the part of the refinery or its contractors. Apparent problems with the October reports illustrate this point quite clearly.
Once again in the October repots, there is a great deal of confusion
over the accuracy
of emissions concentrations being reported in the various data summary tables. On the South UV monitors, for example, we see reported concentrations of carbon disulfide over 400ppb. On the North TDLAS monitors, we see reports of hydrogen sulfide concentrations over 3ppm. In both cases, the project manager at the refinery has told us flatly that he does not believe these numbers to be accurate.
As we have stated earlier, we believe that if the refinery does not believe the emissions information they are publishing in their monthly reports is accurate, the refinery needs to make this point perfectly clear in those reports.
The problem doesn’t seem to be getting any better. To date, (mid-May 1998) we still don’t have any word that the refinery has plans to implement any type of serious data Quality Assurance program for the fenceline monitoring network.
We are not attempting to dispute that some of the reported emissions concentrations in question may be the result of instrument malfunction. We are not really in a position to know one way or another about this. Considering the ambitious nature of this project, technical problems of this sort are understandable, and even somewhat expected in the early going like this. What we cannot understand is the refinery’s failure to get to the bottom of the problems with the monitors.
The monitors at the refinery continue to churn out data. The validity of much of this data seems to be very much in question. In our view, several questions need to be addressed, including:
Why doesn’t the refinery have a comprehensive QA plan for its fenceline monitoring program?
Why haven’t they undertaken traditional canister sampling in the beam paths in an effort to verify the validity of their own data?
Why aren’t they "challenging" all the monitors with reference gases at known quantities at regular intervals, as is recommended by EPA, and as was done during the 6-month test phase of this program?
Presently, the refinery isn’t even attempting to use these traditional methods to validate their data. The integrity of their data is clearly suffering because of this. We find this inexcusable and unacceptable.
In the October data set, as in the two previous months, we were able to re-process only the FTIR spectral data from the North fenceline. The spectra from the South fenceline continue to be of poor quality, apparently due to detector failure(s).
The Operational performance for the FTIR unit on the North continued to be quite good, with an on-line efficiency of more than 98%. As in the previous months, SEA’s analysis of the spectra from the North FTIR reveals that the minimum detection limits (mdl’s) achieved by Tosco’s contractor leave much room for improvement.
Once again, SEA is reporting more than twice as many "hits" of the chemical MTBE on the North fenceline than is being reported by the refinery (see October FTIR data summary). These additional hits are apparently due the better detection limits we have been able to achieve. Our average mdl for MTBE in October on the North is reported at 0.032ppm. Terra’s average mdl for the same compound is reported at 0.071. We see no reason that their detection limits, for all monitored compounds, should not be at least as good as our own.
It needs to be emphasized here that we did not do anything in the analysis of the October spectra which Terra could not have done in their "real-time analysis." The background reference file which was used for the entire month of October was a spectral file generated in early September. This is the same file which we had utilized as a background reference for the entire month of September. We believe that there are several specific steps which Terra could be taking in attempts to optimize their FTIR detection limits. Tosco and Terra appear to be content with the detection limits which are being reported, as we have not seen any significant improvement in these numbers over the past several months.
The refinery will doubtlessly argue that their reported detection limits are "good enough" to provide adequate warning to the community, as the mdl’s they report are still well below any of the pre-set alarm levels. This argument, which seems logical on its face, fails to take in to account the types of problems which can occur with FTIR analysis when the FTIR operator does not make every attempt to optimize real-time detection limits. One of these potential problems is illustrated very clearly in the October data set, as Terra failed to detect propane gas, which was intermittently present on the North fenceline at quantities approaching 2 parts-per-million. (see October FTIR data summary).
During October, we report 34 hits of propane with a maximum concentration of 1.98ppm. Terra reports no detections of propane during the month. It is interesting to note that there are no actual discrepancies reported in the databases. For example, when we are reporting 1.98ppm, Terra’s real-time analysis was reporting non-detect <2.1ppm. While this case does not represent a discrepancy between our reports, Terra clearly should have detected propane at these levels, especially since their average reported mdl for this compound is nearly an order of magnitude lower than this @ 0.283ppm.
As soon as we detected propane, (March 15, 1998) SEA contacted Terra Air Services about the apparent problem. We asked them to confirm the presence of propane in the spectra where we find it, and to attempt to confirm our quantification of this compound. It has been our stated policy that we will attempt to work out any discrepancies with the refinery’s contractor before we report them publicly. To date, Terra Air Services has not responded to our inquiries about the apparent propane hits other than to inform us that they had received our original e-mail message.
In all likelihood, Terra’s failure to detect propane in the spectra in question is the result of a simple and understandable error in the analytical "script" files.(see figure 1 at the end of this report) Because they have not responded to our inquiries, however, we have no way of knowing what the problem is; or even if they agree that there is a problem with their scripts. In any case, it is difficult to imagine that a consulting firm with the experience of Terra Air Services would have failed to detect propane at these levels had they been making any systematic effort to optimize their detection limits!
It comes as a disappointment to us that the refinery and its contractors have chosen not to respond to our inquiries in this matter. We had hoped that they would recognize the problem, fix the problem, and go on from there. Again, at this stage, we have no way of knowing whether or not they even acknowledge that they missed the propane, or if they have taken any steps to keep this from happening again.
We report 152 detections of n-heptane on the North fenceline during the month of October. (see North FTIR data summary) The maximum reported concentration for this compound is 0.178ppm. This is not one of the 39 compounds presently being monitored or reported on by the refinery.
If the refinery’s contractor were to make a legitimate effort to optimize FTIR detection limits, part of this effort would include an attempt to identify any "unknown" or "interfering" chemicals which show themselves in the FTIR spectra, and then to account for these compounds in their analytical "script files". Even if they do not wish or plan to report on the concentrations of these "interfering" compounds, we believe that an attempt should be made to incorporate these chemicals into their analytical scripts.
Once again, we believe that it comes down to a matter of intent. If the contractors at Terra are not being instructed to optimize their detection limits; if they are not actively attempting to identify "unknown" or "interfering" compounds, and to include these as "interferents" in their script files, we believe they will continue to have problems of the sort we have seen with excessive false positive reporting, poor detection limits, and missed detections of potentially hazardous chemicals.
SEA’s concerns about the quality of the data we are receiving from the UV monitors on both fencelines are still among our greatest concerns regarding the operation of the overall fenceline system (please see our September report).
In Terra’s October report, the paragraph dedicated to UV system operation begins with the sentence; "Problems with the UV system have not been fully resolved." We think this is a gross understatement. We do not believe that the primary problem with the UV monitors involves the logging of "excessive downtime".
We at SEA do not believe we are in a position to provide accurate information to the public regarding the operations of the UV system because we do not feel that we are getting straight answers from the refinery to our own questions. We need to know:
Are the emissions concentrations being reported by the refinery accurate?
Has the refinery undertaken any Quality Assurance measures to confirm these reports?
If so, where are the results of the QA testing? If not,
What is Terra referring to when they talk about "differences. . .in baselines for some of the channels" in their October reports?
Are the reported concentrations being measured from "zero", or from some imaginary "baseline" emissions level?
Why are we seeing so many negative concentrations in the databases? Why are these negative values not addressed in Terra’s reports or data summaries?
Obviously, we have more questions than answers about the operation of the UV systems at this time.
For the month of October, Terra reports an "on stream efficiency" of more than 99% for all four of the TDLAS monitors. The reports do indicate that there were some serious problems with the data being recorded by these monitors. Once again, issues of Quality Assurance - or complete lack thereof - are our primary concern.
A review of the TDLAS data summaries for October indicates that Hydrogen Sulfide was detected on the North fenceline at concentrations up to 3.4ppm. In their report, Terra addresses this by saying:
"There were intermittent occurrences of higher than usual H2S results from the north TDLS. Site investigation on several occasions showed these to be false positives."
What is unclear is how Terra determined that these detections were "false positives". Nothing in the provided databases would indicate that the instruments were malfunctioning in any way. All of the "data quality parameters" appear to have been met.
It appears that refinery personnel determined the H2S readings to be false positives based on the fact that they couldn’t smell the chemical, and that no odor complaints had been received from the surrounding communities. We at SEA, as signatory to the Good Neighbor Agreement, find this "sniff test" to be an unacceptable form of Quality Assurance.
We will concede that the refinery may well be right, that the reported concentrations of hydrogen sulfide may have been too high. Once again, our criticism is based on the apparent lack of a serious attempt to confirm this problem via scientific means. If we are asked to rely on someone’s sense of smell to be the final determiner of hydrogen sulfide concentrations at the refinery fenceline, aren’t we right back at square one?
Looking ahead, it is interesting to note that we see very similar H2S concentrations reported in the TDLAS databases for the month of November. Unlike the October data, however, none of these "hits" even make it into Terra’s November Summary Report.
Boreal, the manufacturer of the TDLAS systems, apparently provided the refinery with a "hardware/software" upgrade sometime in February of 1998. This system upgrade is designed to alleviate the "false positive" reporting problems. It is presently unclear whether or not this upgrade has been effective.
The red trace above is an absorbance spectrum from 12:55pm on 10/5/97. The trace below it is a portion of the reference spectrum for propane. Propane is clearly identified in the upper trace. It is easy to see how this incursion of propane could have slipped by the refinery's contractors without being detected - if they were monitoring for this compound in the region surrounding the "strongest absorbing peak" - as water vapor interference is very strong in this region. If this were the case, it would explain the "non-detect" they arrive at, and would also explain the very poor detection limits which they report at this time.
We cannot know for sure if this is what happened here, as neither the refinery nor its contractors have told us what the problem is. They have not even acknowledged that there is a problem, or that they missed this incursion of propane.
SEA does not have a toxicologist on-staff. Except in such cases where a reported emission concentration represents a violation of Federal, State, or local standards, WE WILL NOT BE COMMENTING on the relative risks of any reported release.
By making these reports publicly available, both those emissions figures reported by SEA, and by TOSCO, we do not wish to imply that any of these chemical emissions pose ANY HEALTH THREAT WHATSOEVER to people in the surrounding communities.
SEA member Andy Mechling, who has been retained by SEA to re-analyze the raw emissions data from Tosco’s FTIR monitors, holds no formal scientific credentials. He holds Bachelors degrees in History and Communications. He has undertaken training by the Environmental Technologies Group, the manufacturer of the FTIR instruments, and now has more than two years experience producing and analyzing this type of data for a number of community groups in a wide variety of situations. SEA has every confidence in his abilities to accurately analyze and report this data. Still, considering of our staff’s lack of scientific credentials, any chemical concentration figures reported by SEA should be regarded as our BEST ATTEMPT to arrive at the truth, rather than the ultimate truth regarding chemical emissions from the Tosco refinery.
Because of the nature of infrared spectroscopy, where the raw spectral data is saved electronically, all reported concentrations of chemical emissions are ultimately verifiable. SEA will make the raw spectra which make up our reports available upon request to any party for any reason. Furthermore, in the event that there is a discrepancy between our analytical results and those arrived at by Tosco’s FTIR contractor, SEA will make every attempt to notify the refinery, and to rectify the discrepancy BEFORE making any results publicly available.
We are undertaking this project, and making these reports public, because we believe that residents in our communities have a fundamental right to know as much as possible about the air that surrounds them. It would be much preferable to us if this task were being undertaken instead by one of the regulatory agencies, who possess infinitely better resources than our very small organization. Unfortunately, this is not the case at all, and the job is left for us to do. Simply put, we are making this data public in this fashion because we are presently the ONLY AVAILABLE SOURCE for this type of chemical emissions data.