UST Program - Leak Detection Methods (Leaking Site Survey Report)

• 4.1.1 Examples of the tank tightness test results for the leaking tank cases
• 4.1.2 Examples of SIR results for leaking tank cases
• 4.1.3 Examples of MIR results for leaking tank cases

• 4.2.1 Examples of line tightness test results for leaking line cases
• 4.2.2 Examples of SIR results for leaking line cases
• 4.2.3 Examples of MIR results for leaking line cases
• 4.2.4 Examples of line leak detector test results for leaking line cases

• 5.1 Manual Inventory Reconciliation
• 5.2 Statistical Inventory Reconciliation
• 5.3 Tank Tightness Test
• 5.4 Line Tightness Test
• 5.5 Other Leak Detection Methods

• Table 1 - Summary of Leak Discovery Methods
• Table 2 - Cases Where Leak Detection Discovered The Leak
• Table 3 - Leak Sources
• Table 4 - Cases With Tank as the Reported Leak Source
• Table 5 - Cases With Piping as the Reported Leak Source
• Table 6 - Cases With MIR/SIR Testing Information
• Table 7 - Cases With Tank Tightness Testing Information
• Table 8 - Cases With Piping Testing Information

APPENDICES

• APPENDIX I - LEAK SITE SURVEY LETTER
• APPENDIX II - LEAKING SITES SURVEY - LOCAL AGENCY RESPONSE SUMMARY
• APPENDIX III.A - COMPLETE DATABASE FOR THE LEAK SURVEY
• APPENDIX III.B - LIST OF DATABASE ABBREVIATIONS AND DEFINITIONS

The increased use of oxygenates as fuel additives combined with recent concern over environmental and health effects of some of these additives, such as MTBE, has drawn attention to the reliability of leak detection systems. A survey was conducted to determine the effectiveness of leak detection methods in finding leaks from underground storage tank (UST) systems in California. This report is the result of the survey which includes information about 345 leak cases reported between October 1995 and May 1996. In 313 cases we have information on how the leak was discovered. In 84% of the cases, the leak was discovered during closure.

The survey data validates the concern that most of the leaking UST sites are not consistently monitored. In 281 cases, cases for which local agencies responded to the request for additional information, an estimated 149 sites (53%) were not monitored (a few because the tanks were abandoned), monitoring records were not in the local agency files, or monitoring history was not known. For the 132 cases with available monitoring information, there are long time gaps (an average of 29 months) between the most recent monitoring report and the discovery of the leak. This time gap makes evaluation of leak detection methods difficult. Although the information about these sites is not complete, this report highlights some major concerns about the use of leak detection. Analysis of this data emphasizes the need for more rigorous surveys on the use of leak detection methods and their field performance. This report also helps to identify additional data needs and strategy of data collection for future surveys.

Leak detection methods correctly identified leaking tanks or piping in approximately 4.8% (15/313) cases. So it appears that if used properly, leak detection has the potential to identify leaks. Isolated cases in the data highlight two of the main concerns with leak detection. These concerns come from ignoring or overruling (by performing another test) the failed monitoring results and incorrect reporting of monitoring results. In some cases it appears that the leak detection method failed to identify the leak, because the leak was discovered shortly after the last monitoring report indicating a tight system.

Most leaks were from tanks and piping systems. Of the 121 cases for which leak source information is available, 50% (60/121) were tank leaks and 34% (41/121) were piping leaks. A total of 18 dispenser area and 12 overfill/spill leaks were reported. Most of the leaking systems were single-walled USTs from 10 to 40 years old. Since January 1, 1984, state regulations have required that all new tanks have secondary containment (for piping this has been required for all tank systems installed after July 1, 1987). There are about 10 cases where the leak most likely came from a double-walled tank system (less than ten years old). One leak was due to a fiberglass tank rupture (discovered during the annual equipment maintenance check), and two of the leaks were in the dispenser area.

Due to the lack of available records, this survey does not include information about Automatic Tank Gauging (ATG) systems, ground water and vapor monitoring and Manual Tank Gauging (MTG) methods. Although there is an indication that some of these sites have been using ATG systems for monthly monitoring, none of the leaks in this survey were discovered by ATG systems.

An aggressive enforcement of leak detection requirements and another more comprehensive survey of leak detection methods with extended resources for collecting complete site-specific information is recommended.

The purpose of this study was to determine if leak detection methods are finding leaks in leaking underground storage tank UST systems in California. A database representing all the leaking UST sites reported between October 1995 and May 1996 (345 cases) was constructed using MS ACCESS software. This information was obtained from the Leaking Underground Storage Tank Information System (LUSTIS) database which contains information reported in the "UST Release Report" forms. A survey of local agencies was conducted to obtain monitoring information for these sites.

On June 14, 1996, a letter requesting monitoring information for the leak cases from each local agency jurisdiction was sent to each agency (see sample letter in Appendix I). Of the 60 local agencies surveyed, 42 responded. Appendix II is a summary of the number of cases in each jurisdiction, a list of the local agencies that responded to the survey, and those that did not. A response from the local agencies was received for 281 cases (81% of the total number of cases in the database). Responses ranged from "no additional information available" to providing detailed monitoring records including tank and piping test reports.

The information from these responses was merged with the relevant fields of data from the LUSTIS into a new MS ACCESS database. A copy of the database information (excluding site specific details) and description of the database abbreviations are included in Appendix III .

In the original LUSTIS database, the "How Discovered" field included six possible responses: inventory reconciliation, subsurface monitoring, tank tightness test, tank closure, nuisance conditions, and other. Many cases did not include any information on how the leaks were discovered. In cases where the agency provided additional information, the "How discovered" field was completed and/or modified. The new database includes 12 possible responses to "How discovered." A summary of the number of cases for each category is included in Table 1.

These data indicate that most leaks are discovered during tank closure and removal, not by leak detection. About 263 leaking UST systems were discovered during tank closure/removal or other site-work related activities. A total of 15 cases were discovered by a leak detection method (see Table 2 for the breakdown).

The leak detection method reporting the highest number of "fail" results was tank tightness testing. But one should note that the majority of available leak detection information in this survey was the tank and piping test results. Two piping leaks were discovered by inventory reconciliation, one by a line tightness test, and one by mechanical line leak detector going into slow flow (the electronic line leak detector was disconnected). Tank tightness testing discovered eight leaks, six with source of the leak "unknown" (two of these cases failed both a tank tightness test and a line tightness test) and two with piping as the leak source. No information on the type of tank tightness test method that discovered piping leaks is provided. Annual leak detection equipment checks discovered two leaks. One leak was due to a double-wall fiberglass tank rupture the source of the other leak is not listed.

4.0 REVIEW OF DATA BASED ON LEAK SOURCE

One should note that often the leak source information is the best guess by the person completing the release report form or by the inspector on site. Without detailed review of the analytical results, it is not readily possible to identify the leak source. Even with soil analytical results, locating the leak source may not be possible because the sampling locations during the site investigation are designed to determine presence or absence of contamination and the extent of contamination.

Table 3 provides a summary of the information on the leak sources of the UST systems for this survey. For 224 of the cases no information on the leak source was available (no response or no records). Of the remaining 121 cases, 60 reported the tank as the leak source, 41 were piping leaks, 18 were in the dispenser area, and 12 involved spill and/or overfill. Past leaking UST site surveys indicate that most leaks are from piping and spill/overfill. This database indicates that the majority of the leaks are from tanks and then from piping. Interestingly, the number of leaks due to spill/overfill is very low, 10% (12/121). This could be attributed to the increased use of spill/overfill protection equipment. Dispenser area leaks account for 15% (18/121) of these leaks.

In most cases, the age of the UST system is unknown. Of those with available tank age information (116), the age of the majority ranged from 10 to 40 years. These must be predominantly, that is mostly old single-wall systems, because in California single-wall tanks have not been allowed since 1984. There were about 10 cases for which the estimated age of the UST systems were less than ten years, indicating that most likely they were new double-walled systems. For one of these cases the source of leak was a rupture in the tank, which was discovered during the annual leak detection equipment maintenance check (required in the State of California). The source of the leak for the rest of these sites were dispenser area (two cases), turbine connections (one case), a day tank leak (one case), pre-existing contamination, and some not reported.

The information for the cases where the tank is the reported leak source are tabulated in Table 4. The majority of the 60 tank leaks were discovered during tank closure activities (tank removal, site assessment, soil boring, subsurface monitoring) and none reported by tank tightness testing. For four of the sites, "other means" was marked in the "how discovered" column of the leak report form. At one site the leak from a double-wall tank was discovered during the annual leak detection equipment check (required per California regulations). This leak was due to the rupture of a fiberglass tank.

Following is a summary of analysis of the information presented in Table 4:

• There were 60 sites with tanks as the reported leak source.
• Limited monitoring results were available for 17 of these 60 sites. One site was monitored by an interstitial monitoring device (it was a double-wall tank). Fourteen of the sites had reported tank tightness test results.
• Of these leaking tank sites, 35 were not monitored or did not report monitoring results to the local agency. Out of these 35 cases, 13 of them were at sites with abandoned or exempt tanks. It is possible that some of these sites may have been using ATGs for performing monthly tank monitoring. State regulations do not require submittal of the monthly ATG test results to the local agency. However, like other leak detection methods, any "fail" test results must have been reported to the local agency. So even if some of these sites had ATG systems installed, no failures were reported.
• None of the tank leaks were discovered by tank tightness testing or other monitoring methods. In one of the cases local agency reported that the tank tightness test results were suspected of being falsified.
• The last available tank tightness test report for the 14 cases, that had a test result available, reported a "pass". Two of the 14 cases were from sites where tanks had previous failed tank tightness test results.
• The time lapse between the last tank tightness test and discovery of the leak ranges from 6 months to about 5 years. One test was performed within 6 months of the leak discovery date, the rest were done more than 20 months prior to the leak discovery date. For the case that a tank tightness test was performed six months prior to the leak discovery date, information on the age of the leak would have been helpful in determining if a leak was present during the tank tightness test.
• For one of the leaking tank sites, SIR was used as the monthly monitoring method. The last SIR analysis performed for the period five weeks prior to the leak discovery date reported a "pass". However, there were previous "excess" gain SIR test results that were reported a "pass." The last available tank tightness test data for this site was 33 months prior to the leak discovery date (note that tank tightness test is required every two years in conjunction with SIR).
• Four of the leaking tank cases were at sites using monthly MIR monitoring and one was at a site using MTG. According to the agency’s records, no gains or losses outside the allowable range were reported. For all of these five sites, the time lapse between the last available annual MIR summary report and the leak discovery date is a year or more. These are good examples of tank owners failure to perform leak detection and/or to submit required monitoring reports to the local agencies.
• Age of the tanks at these sites ranged from 7 years (double-wall) to 38 years, but for most of the cases it was unknown. It appears that most of the leaking USTs are single-wall systems. As mentioned earlier, a total of 13 of the leaking cases were abandoned or exempt tanks (farm or residential tanks).

We cannot objectively report on the ability or failure of the tank tightness testing method due to the lack of consistent tank tightness test history, the long time gaps between last tank tightness test and leak discovery date, and the unknown age of the release. However, the lack of proper monitoring at many of these sites is noted. For future surveys, information on the estimated age of the release should be collected when studying the effectiveness of leak detection methods that are used less frequently (i.e. annual or every two years).

At a Fresno site, a tank tightness test was conducted six months prior to discovery of a leak from a 550 gallon regular unleaded tank. The calculated leak rate for this test was reported as 0.039 gallons per hour (gph), which is less than the 0.05 gph threshold, but the test report does not include enough detail to determine if the tester followed the test protocol properly.

At two of the sites, there were old failed tank tightness tests that were followed by passing tests. For these two sites there is no record to indicate that an investigation was done due to the failed test. One is a site in Compton for which contamination was discovered during tank closure. Tank tightness testing performed on 10/16/85 failed the gasoline tank (one of the 5 tanks on the site, there was also a clarifier tank on site that was never tested). However, that tank passed the tightness test during the 3/19/86 test. There are no records of any action taken as a result of the 1985 failed test result. The diesel tank at the same site passed the tank tightness test on 10/16/85. No records are available to indicate if the lines were tested. According to the local agency notes the source of the leak at this site was piping and dispensers.

The other case is a site with two tanks that both failed the tightness test performed on 3/11/93 and the re-test on 3/25/93, but passed the third test on 4/2/93 (apparently the same tester using a different equipment ). There is no information of any follow-up actions due to the initial "fail" tests. Both tanks passed the tightness test on 3/16/94. According to the response from the local agency, this site was using MIR monitoring method, no records were provided.

For a Miramonte site, the tank was recorded as the leak source, but according to the local agency the tank appeared to be sound when it was removed. In case of a site in Guinda, the source of the contamination was a tank with no permit. At a Alhambra site, a 1988 site investigation indicated a leak. However, the date of discovering the leak in the LUSTIS database was recorded as 3/6/91. Comparing the date of last tank tightness test (1/28/87) with the 1988 time line, the leaking tank had passed a tightness test conducted a year prior to discovery of the problem.

These examples indicate that "fail" results tend to be ignored or overruled. Also with no available records indicating correction of a problem that may have led to the "fail" test result, one suspects that the follow-up "pass" test results may not be correct.

There is only one case with the tank as the leak source, for which provided information indicates that SIR monitoring was used (Kelseyville site). For this site, the last monthly SIR test report available is the 10/31/95 report, which is 5 weeks prior to the date the leak was discovered (12/6/95). A review of the annual SIR summary report indicates that test results for one of the tanks showed gains higher than the leak threshold for the 2/94, 8/94, and 9/94 period, but these gains were reported as "pass" by the vendor. The annual summary report does not specify the name of the SIR vendor. The last tank tightness test records were from 3/31/93 and 7/1/94, another example of overdue tank tightness test.

There is not enough SIR evaluation data available to draw a conclusion on the performance of the method. However, available information indicates that ignoring "gains" could lead to missing a tank leak. "Gains" must be treated the same as losses when reporting monitoring results.

At five of the leaking tank sites monthly inventory reconciliation was the monitoring method and at one site monthly MTG was listed. However, annual summary MIR or MTG reports were generally not available. According to the California regulations tank owners are required to submit a summary of the MIR and MTG results to the local agencies on an annual basis. For the cases with some annual MIR (or MTG) summary reports, the date of the last report is more than a year from the date of leak discovery, so it is difficult to evaluate ability of these methods in identifying problems. All the available annual MIR results for these sites have been reported as passes (within allowable range).

The information for the cases where piping is the reported source of leak is tabulated in Table 5. Review of this data suggests the following observations:

The leak source for 42 of the cases was piping. For 38 of these cases a response from the local agency was received. Only five of the cases with the line as the leak source had line tightness test results available. The remaining sites either did not have a piping test performed or did not provide the report to the local agency. The last line tightness testing for the five sites were performed zero to ten months prior to the leak discovery date. One of the leaks was discovered as the result of line tightness test. One leak was discovered due to the loss of pump prime. Only one leak was discovered by the line leak detector. It is not known if working line leak detectors were present at the pressurized lines. The information on the type of piping (pressure or suction) was also not always provided. For 13 of the leaking piping sites, tank tightness test results were reported. For two of the line leak cases, "tank tightness testing" was the reported leak discovery method. There was no information on the type of the tank tightness test method and on whether a piping test was also performed. For three sites MIR records of monitoring and for one site SIR records of monitoring were provided. In one case the local agency has reported MIR as the leak discovery method. In this case local agency response indicates that the result of the last line tightness test was a "fail". It appears that tank and line tightness test were performed to verify the problem.

Out of the five cases for which piping tightness test results were available, two were discovered by line tightness test and the other three had reported a "pass" for the last line tightness test. These "pass" test results were performed two months, nine months, and ten months prior to the leak discovery date. Most of the piping leaks were discovered during tank closure and removal or other site work. In the case where the line tightness test passed the line nine months prior to discovery of the leak, a mechanical line leak detector lead to leak discovery by triggering slow flow (electronic line leak detector was also installed but it was disconnected). Mechanical line leak detectors are designed to detect major line leaks, leak rate of 3 gph or higher. So it appears that this line leak was not a slow gradual release, and therefore, the line may have been tight when the line tightness test was performed. This emphasizes that availability of reliable information on the age of the release is important to an objective interpretation of tank and line tightness test results.

SIR records of monitoring were only available for one of the sites with a reported line leak. At this site a loose union was found to be the cause of the leak, which was discovered during the line tightness test (10/24/95). The last SIR report prior to discovery of this problem was for the months of 9/95 and 10/95, and they reported a tight system. There were three "fail" SIR reports for each of the tank systems at the site during 1994. Vendor attributed the "fail" test results to the errors in dip stick reading. There are no records of further work to confirm the reason for the "fail" reports. California’s regulations do not recognize SIR method as a monitoring option for piping.

MIR records of monitoring were available for four of the sites with a reported line leak. At 3 of these sites, there were no reported variations outside the allowable range. At one site, MIR monitoring lead to the discovery of the line leak which was also confirmed by a line tightness test. Generally, MIR is not accurate enough to detect minor line leaks (below 1 gph) even when performed correctly and consistently.

One line leak, out of the 42 reported line leak cases, was discovered by a line leak detector (LLD). According to the UST regulations all single-wall pressurized piping systems must be equipped with LLDs. The reason for failure of these systems in detecting the line leaks may have been one or combination of the following factors: no LLDs were installed, they failed to detect the leak due to improper installation or a defect, leaks were less than 3 gph (leak detection criteria for an hourly line test is 3 gph), the LLDs were disconnected and not operational, or they detected the leaks but they were by-passed.

Dispenser area leaks were reported at 18 sites. At two of these sites, available information indicates presence of suction system (a suction pump is located under the dispenser). Most of these cases were discovered during tank closure activities, a few cases were discovered during dispenser relocation.

At 11 of these sites, a tank tightness test and for 7 of them piping tightness test were performed within a year or longer time period prior to discovery of the leak. A few of the tank tightness test methods were volumetric overfill test methods and some included ullage tests. In one case it was reported that the site had an ATG (these systems are currently not designed to detect leaks from the piping. MIR monitoring was in place at eight of these sites. For one of these cases MIR variations over the allowable range were noted during the year of leak discovery.

Generally, dispenser area leaks are from under the dispenser piping, suction pumps, or due to poor filter change practices. In some cases customers that overfill their gas tank also contribute to dispenser area contamination. The only effective way of discovering under the dispenser leaks, when there is no secondary containment, is frequent visual monitoring. For major leaks, inventory type test methods may also detect under the dispenser piping leaks. However, it is best to prevent under the dispenser contamination by the installation of dispenser containment.

Of the 281 cases, for which local agencies responded to the survey, leak detection information was available for 132 sites. Out of 313 sites for which leak discovery method information was in available, 15 leaks were discovered by leak detection. At most of these sites monitoring was not performed consistently.

No monitoring results on the use of ATG, ground water and vapor monitoring was available for the subject sites. Therefore, the analysis in this report excludes an evaluation of the effectiveness of these leak detection methods. However, one should note that none of the cases for which "leak discovery" information was available had reported any of these three monitoring methods as the method of discovering the leak.

Some MIR information was provided for 20 of the leaking sites. For none of these sites there is an indication that the tank owner ever reported a discrepancy (inventory variations outside the allowable range). For nine of these cases the dispenser was the leak source (for one case tank leak and for another one overfill also contributed to the contamination), three were spill and overfills, five were tank leaks, and two were piping leaks. For the remainder of the cases, the leak source was not reported. The time frame between the last reported MIR result and the leak discovery date ranges from one to six years.

For two of the cases, the agency reported that the MIR method of leak detection lead to discovery of the leak, both cases were piping leaks. This indicates that, if performed consistently and properly, MIR has the potential to detect leaks from the UST system (generally major losses). The dispenser area leaks are generally gradual leaks at a rate below the detection range of MIR, which at its best can detect a leak rate of one gallon per hour (about 720 gallons per month) or higher. Information on the extent and age of release would be useful for future surveys of this kind.

The major problem with the use of the MIR monitoring method appears to be the tank owner/operator failure to do it or to report the results. It is also unusual for all the MIR results throughout the year to be within allowable variations, that means no false alarms. This method is very prone to errors (calculations, dip stick errors, chart conversions, etc.).

The MIR method of monitoring will be phased out after December 22, 1998. This method requires detailed record keeping, diligence and consistency. The other problem with the use of this method for monitoring UST systems is that even when performed properly, it could still miss leaks as large as 720 gallons per month.

In this survey, SIR information was only provided for a total of nine sites. This is not adequate data for an objective evaluation of the effectiveness of SIR analysis. The time period between discovery of the leak and the last passing SIR report for these few sites ranged from zero to 23 months. None of the leaks at these sites were reported by the monthly SIR analysis.

The source of the leak for one of the nine sites was the piping and for another one was the tank, for the remaining cases leak source is not reported. The piping leak was discovered by line tightness test and it was caused by a loose union. The SIR result for the month during which this leak was discovered was reported as a "pass." There were three "fail" SIR reports during the year prior to leak discovery on all the tank systems at this site. They were attributed to "dip stick errors." The tank leak was discovered during tank closure. The last passing SIR report (10/31/95) was for the period one month prior to the discovery of the leak (12/6/95). SIR results for this site were available for the two years up to the leak discovery month (from 2/94 through 10/31/95) . During the year prior to the discovery of leak, there were at least three SIR results where the analysis indicated excess gains, reported as "pass" and ignored. The last tank tightness test result available for this site was conducted 3/31/93, two years and nine months prior to the leak discovery date. In the State of California, tank tightness test, conducted every two years, is the required supplement for monthly SIR monitoring.

Two of the leaks could have potentially been detected by use of the SIR method. In one case, the SIR vendor reported a "pass" for the month during which a leak was discovered. This leak was discovered during the tank upgrade assessment work and the source of leak for this case was not reported. Unless the leak source was spill/overfill, then SIR analysis should have reported the problem. The other case is a tank leak where the SIR result for the month prior to the leak discovery date was reported as a "pass." Contamination was found during tank closure.

This survey does not provide an adequate number of cases to allow a complete evaluation of the effectiveness of the SIR method of leak detection. However, it does suggest the need for improving SIR reporting practices (especially reporting of "excess gains"), better quality control in the data collection, and better attention by the tank owners and local regulators to the reported "fail" and "inconclusive" results. Most of the leaking sites in this survey have had several inconclusive, fail, and excess gain results (generally reported as a "pass" or "inconclusive"). Apparently, most "fail" and "inconclusive" reports are considered to be the nature of this type of monitoring instead of leading to a concern with a potential leak.

Table 7 provides a summary of cases for which tank tightness test results were reported. There are 120 cases in the database for which tank tightness testing results are available. For 14 of these cases the tank was the reported leak source. None of these tank leaks were discovered by tank tightness testing. However, a total of eight leaks were discovered as the result of tank tightness testing, two had reported piping as the leak source and the remaining leak sources not reported. The date of last available tank tightness test result and the leak discovery for these 120 cases range from six months to five years.

Tank tightness testing is required annually when the monthly monitoring method for the tank is MIR, and it is required every two years when SIR is the selected monthly monitoring method. Tank tightness test methods are third-party certified for detecting tank leaks, but when overfilled type tests or vacuum/pressure test methods are used, piping leaks may also be detected. Tank testers also note and report the loose piping connections during the site preparation for a tank tightness test.

The major problem with use of tank tightness testing for leak detection appears to be the tank owner/operator failure to have it performed and their ability to continue to operate their tank without compliance with this requirement. Another concern with tank tightness testing, when conducted, is that the "fail" test results are ignored or overruled. In some instances "fail" tank tightness test results were followed by a "pass" report using another test method or by calling another tank tester. A few of the reported leak sites with a "pass" on their last tank tightness test have had history of unexplained "fail" test reports.

The main problem with the evaluation of effectiveness of tank tightness testing in detecting leaks is that it is not a monthly monitoring method. So, unless the leaks are discovered in a time frame reasonably close to the time the test was conducted, it is difficult to conclude that the method failed to detect the leak since the tank may have been tight at the time of last test. Information on the estimated age of the release is important when evaluating effectiveness of this type of leak detection methods.

During the review of the test reports, we also noted that some testers did not follow the requirements of the test protocol properly, or did not provide detailed enough report to determine accuracy of the test. In one case the local agency reported that the tank tightness test results were suspected of being falsified.

Table 8 provides a summary of cases for which line tightness results are available. There are 46 cases with reported line tightness test results. Five of these cases had reported the lines as the leak source and 8 had reported the dispenser area as the leak source. One of the line leaks was discovered by line tightness testing, none of the dispenser area leaks were reported as a result of line tightness testing. The time lapse between the last available line tightness test result and leak discovery ranged form zero to ten months. Line tightness test is required on pressurized lines annually and on suction systems every three years.

The major problem with use of line tightness test as a leak detection method appears to be the tank owners/operators failure to have it performed at the required frequencies and lack of adequate enforcement of this requirement. The main difficulty with the evaluation of effectiveness of line tightness test again is the testing frequency. Unless the last line tightness test date and leak discovery date are reasonably close, without information on the age of the leak, it is difficult to determine if the system was actually leaking during the test.

Annual maintenance check on leak detection equipment and mechanical line leak detectors have identified problems in a few cases. There is not enough data to evaluate effectiveness of LLDs in detecting leaks.

This database has many deficiencies that make effective use of it difficult. However, it helps identify the need for future work on determining the effectiveness of leak detection. It also makes it clear that detailed records are needed as well as the cooperation of various groups to build a useful database.

Overall, it appears that the leak detection methods are not discovering many leaks. However, the main problem appears to be that many of the leaking sites have not been properly monitored. The following summarizes the conclusions of our analysis:

• Most of the leaking sites are old single-walled facilities.
• Most leaks are discovered during tank closure/removal or other site activities.
• Most of the leaking sites in the database had not been monitored at all or were not monitored consistently.
• There are some cases where it appears that leak detection method may have missed detecting a leak.
• Some tank testers do not provide adequate test reports and some fail to follow the manufacturers’ protocol.
• The reported "fail" and "inconclusive" reports by SIR vendors and tank testers seem to be ignored by tank owners, not investigated thoroughly, or overruled by another test.
• Some SIR vendors do not provide adequate test reports to allow review of the results and/or do not report test results correctly. The most ignored or improperly reported results are "excess gains".
• Operators using the MIR method of monitoring do not perform it consistently and in many instances do not provide annual summary statements to the local agencies, as required by the regulations.
• The effectiveness of monitoring methods used less frequently than monthly is difficult to determine without information on the estimated age of the leak. This in turn suggests that use of frequent monitoring methods over annual type monitoring are the preferred alternative because of their ability to detect leaks within a reasonable time frame.

• Local agencies should enforce consistent use of monitoring methods and record keeping requirements, take enforcement action against tank owners and operators who fail to comply, and enforce annual maintenance check of the leak detection equipment.
• Local agencies should review submitted monitoring records for accuracy, and require proper investigation and documentation of "fail" and "inconclusive" monitoring results.
• Another more detailed survey should be conducted. As necessary, tank owners should be contacted, sites visited, and local agency records reviewed to gather relevant monitoring information. Future surveys should also include ATG test results.
• A study based on information collected during future tank closures should also be conducted. This survey should gather all the relevant contamination and monitoring information to allow an objective evaluation of effectiveness of leak detection methods. Additional data needs are: better identification of the leak source (in case of tanks identify which tank), age of the leak, extent of the release (high versus low, deep versus surface contamination.
• Use of dispenser containment to prevent under-dispenser contamination should be required. Existing leak detection methods are not designed to provide early warning of dispenser area leaks, unless there is a dispenser containment with a monitoring device.