Tritium Releases and Potential Offsite Exposures
LAWRENCE LIVERMORE NATIONAL LABORATORY (U.S. DOE)
[a/k/a LAWRENCE LIVERMORE NATIONAL LABORATORY (USDOE)]
LIVERMORE, ALAMEDA COUNTY, CALIFORNIA
EPA FACILITY ID: CA2890012584
LAWRENCE LIVERMORE NATIONAL LABORATORY (U.S. DOE)
[a/k/a LAWRENCE LIVERMORE NATIONAL LABORATORY (USDOE)]
LIVERMORE, ALAMEDA AND JOAQUIN COUNTIES, CALIFORNIA
EPA FACILITY ID: CA2890090002
THE SAVANNAH RIVER SITE (U.S. DOE)
[a/k/a SAVANNAH RIVER SITE (USDOE))]
AIKEN, AIKEN, BARNWELL AND ALLENDALE COUNTIES, SOUTH CAROLINA
EPA FACILITY ID: SC1890008989
March 11, 2002
The Lawrence Livermore National Laboratory (LLNL) was established in 1952 for the principal purpose of conducting nuclear weapons research and development. It occupies a 3.3 km2 site next to the City of Livermore, California, and about 60 km east of San Francisco (See Figure 4-1). More than 6 million people live within 80 km of LLNL, about 73,600 of them in the City of Livermore. LLNL also operates a test site near Tracy, California, called Site 300, which has been used principally for testing non-nuclear explosives. Tritium has been used at LLNL in the nuclear weapons research and development program. Releases to on-site and off-site environments have been measured and the methods and results documented in the LLNL Environmental Monitoring Plan (Tate et al. 1999) and Annual Environmental Reports (see, for example, Harrach et al. 1998, Larson et al. 1999, Larson et al. 2000).
The current sources of tritium from the LLNL are the Tritium Facility (airborne releases), several diffuse sources involving mostly storage of tritium-contaminated wastes and materials (airborne, surface water and groundwater releases), and sewage discharge.
The major source of airborne tritium emissions is Building 331 (the Tritium Facility) where experiments involving radioactive gases and their compounds are carried out. There are two stacks, 30 m high which are continuously monitored with two Overhof ionization chambers, one on high range, the other on low. The ionization chambers have alarms and a signal that indicates the total activity that is emitted (HT+HTO plus any other gaseous tritium radioactive compounds). The response of the ionization chambers and alarm set points are verified weekly with a check source. There are quarterly checks of the monitor electronics and a semi-annual check of the chamber calibrations. The monitors are equipped with emergency power supplies. Their current function is primarily for monitoring episodic releases rather than smaller day-to-day routine releases. In the past, when there were significantly higher emissions, the monitors were able to measure routine releases.
In parallel with the ionization chambers on each stack are pairs of molecular sieve continuous samplers in series. The upstream sampler collects HTO; for the downstream sampler, a palladium catalyst converts any HT to HTO so that from the measures of tritium in the moisture collected by the respective samplers, both HTO and HT activities in the effluent can be deduced. These samplers are collected weekly for subsequent extraction of the moisture and measurement of the tritium by liquid scintillation counting. These values are used for reporting emissions and as input to calculations with dispersion models. Other parameters needed for the dispersion models are known (e.g., stack heights) or are continuously measured (stack flow rates, wind speed, direction, and fluctuation characteristics).
There are smaller emissions from several other facilities (for example, Building 391, Laser Target R&D) and four areas that are diffuse sources of tritium on the LLNL site. The diffuse sources are:
Near Building 292-- past leakage from an underground tank with tritium-contaminated water;
Near Building 331-- a waste accumulation area where tritium-contaminated equipment is temporarily stored;
Building 514 -- tanks used to store and treat liquid and solid radioactive and mixed wastes that often contain tritium; and
Near Building 612 -- several outdoor containers holding tritiated wastes.
There has been surveillance of ambient air since the 1970s. Currently, HTO vapor on the Livermore site is monitored with an array of 11 continuously operating silica-gel samplers with an additional six samplers in the Livermore Valley. The samplers are changed every two weeks. Duplicate samplers for quality control, also sampled every two weeks, are operated for two months in parallel with the permanent sampler at a given site.
The Experimental Test Site (Site 300), 34 km to the East of LLNL, is a further diffuse source of tritium emissions from past explosions involving tritium and from tritium-contaminated material disposed in the Site's landfills. Based on a study carried out in 1996 and 1997, which indicated HTO concentrations in air close to general background, only one sampler for HTO is currently maintained on Site 300.
The approximately one million liters of sewage per day that LLNL (plus Sandia National Laboratories) discharges through Building 196 to the city of Livermore collection system is monitored for tritium. Daily and weekly flow-proportional composite samples are collected and analyzed.
Storm water and rainwater are monitored for tritium. Stormwater runoff has been monitored since 1991, and samples are now collected from nine locations on or close to the LLNL site and at eight locations at the Site 300. Samples are taken a few times per year depending on the location and conditions. Rainwater has been collected and analyzed since 1971 and is currently sampled at 15 locations on the LLNL site and in the Livermore Valley.
Tritium in groundwater down gradient of the LLNL site has been monitored since 1988. Currently, groundwater down gradient from the Taxi Strip Area (location of four liquid waste disposal pits that date from 1953 to 1976) and the East Traffic Circle is monitored. A network of 21 wells on or near the LLNL site (including 12 wells in the Livermore Valley) is sampled at least semi-annually. At Site 300 there are several tritium-bearing groundwater plumes (mostly from pits in which tritium-contaminated materials from firing tables are buried) that are monitored. Many wells are sampled and tritium levels determined on and just off the site.
Cores of sediments from storm drainage channels and the two arroyos on the LLNL site are collected for tritium analysis by freeze-drying to recover water from the samples and subsequent LS counting.
LLNL has monitored tritium in vegetation to some extent since 1966, and more extensively (as part of the continuous monitoring program at the LLNL site and Site 300) since 1972. The water in the samples is collected using freeze-drying techniques, and the tritium content of the water then determined by liquid scintillation counting. Vegetation, usually grasses and weeds, are currently sampled quarterly from the Livermore Valley, from San Joaquin County, and from Site 300. The sampling locations are chosen to provide vegetation from places that (1) may be contaminated by tritium from LLNL (within 1 km of the site boundary), (2) background locations where vegetation is similar to that growing near LLNL but is unlikely to be affected by LLNL operations, and (3) areas of known or suspected LLNL-induced contamination. In addition samples are collected from two locations within, or near, the boundary of Site 300 where tritium is known to be present in the subsurface soil.
Locally available wine has been monitored since 1977. The samples have included wines produced from varieties of grapes grown in the Livermore Valley, wines produced from grapes grown in six other wine-producing areas in California outside the Livermore Valley, and wines produced from grapes grown in four areas in France, Germany and Italy. Tritium concentrations are determined by 3He mass spectrometry, a method much more sensitive than liquid scintillation counting and required to measure the tritium activity at the low concentrations found in the wines.
The monitoring methods employed for measuring tritium in effluents are those commonly employed elsewhere and should be capable of providing accurate measures of total gaseous tritium releases (taken by ionization chambers) and of HTO and HT (taken by the molecular sieve samplers). Detailed operational protocols and quality assurance procedures have been developed.
The results from the molecular sieve samplers have been reported for compliance purposes. Ideally comparison of these estimates of releases with time-integrated activities measured with the low-range ionization chambers would indicate whether there might be any tritium in gaseous forms other than HTO and HT (for example, as methane or as larger volatile organic compounds) released in activities that are not negligible compared with those as HTO and HT. The release rates appear to be too close to, or below, the detection limit for the ionization chambers for this direct comparison. Were there to be any episodes of higher release rates, such a comparison would be worthwhile.
The air monitoring with silica gel samplers appears appropriate. The protocols that were sampled (Tate 1998, Tate 1999, Hunt 2000) are detailed and straightforward. The distribution of the sampling locations appears to have been developed on the basis of knowledge of potential source locations. Tritium as HTO is the only chemical species of tritium that is measured. The current program appears adequate and, as with the entire environmental monitoring program, is subject to review annually with updates every three years.
Sewage, groundwater, surface water, and sediments are all currently measured for HTO at locations that have been judged appropriate based on knowledge of piping systems, hydrogeology, and the results of surveys. Given the detailed nature of the hydrogeological work as reported, it seems unlikely that significant plumes of tritium-contaminated water have escaped detection. Storm water runoff was not monitored before 1975, so releases by this pathway are not known for those earlier years. The observed tritium groundwater concentrations reflect much earlier releases from on-site sources, given the slower- moving nature of aquifers.
The tritiated water content of vegetation has been sampled for the past 34 years; more extensively for the past 28 years. The choice of grasses appears to be consistent with the U.S. Department of Energy guideline (USDOE 1991) on sampling indicator materials. Apart from a continuing series of measurements on a pine tree and on a variety of wines, the results from the native grasses appear to provide (at least as reported in the annual reports) the only data on tritium (as HTO) currently in vegetation. Given the low levels observed this might be justified. The freeze-dry analysis method used for both vegetation samples and for air moisture samples with low levels of tritium, as described (LLNL 1999a), is a standard and reliable technique.
The only measurements of tritium that include OBT along with HTO, other than the ionization chambers stack monitors on Building 331, are those on wine. The technique described (LLNL 2000) ensures that total tritium is measured. The only pre-processing of the wine samples is to cycle them through a few heating and cooling steps to remove 3He that may have grown-in before the analysis starts. Hence, the water and solids of the wines are not separated so that the measured 3He will have grown-in from total tritium, i.e., it will be from both HTO and OBT. Helium-3 mass spectrometry would have to be used if any attempt were to be made to measure OBT and HTO separately in vegetation.
Overall, the technologies currently employed in the monitoring and measurement of tritium appear to be consistent with practices described in the literature (NCRP 1976, Wood et al.1993) and widely practiced in nuclear establishments elsewhere.
Most released tritium has been airborne with the dominant source being the Tritium Facility (Building 331). The total activity of tritium released from LLNL has been reported as approximately 28 PBq (760 kCi) during LLNL's more than 40-year history (Tate et al., 1999). Most of this activity was released in two events; 13 PBq (350 kCi) in 1965 and 11 PBq (300 kCi) in 1970. Both releases were cited (Myers et al. 1973 and USDOE 1982) as being in the form HT. Annual reports indicate that releases measured since the 1980's from the Tritium Facility have declined (see Figure 4.2); the large decrease since 1992 reflecting substantial reductions in operations at the tritium facility. The annual tritium releases from LLNL were between 40 to 280 TBq ( 1 kCi to 7.5 kCi) before 1992, and have been between 3 and 12 TBq (80 and 320 Ci) since 1992. During 1999 about 8 TBq (200 Ci) of HTO were emitted from the stacks, about 0.4 TBq (10 Ci) from all other diffuse sources combined, and less than 0.01 TBq (0.3 Ci) were released through the sewer. Releases from other (un-monitored) facilities estimated either from the air surveillance measurements or by using a specified EPA fractional release standard, will add to the measured emissions.
According to the respective annual reports, approximately 80% of the tritium released from the Tritium Facility in the last five years has been in the form of HTO. Reported releases in TBq in each of these years as HTO and as HT respectively have been(5): in 1999 (7.9, 2.5); 1998 (3.1, 0.9); 1997 (9.9, 1.2); 1996 (6.7, 1.2); and in 1995 (2.3, 1.1). Given the much greater dosimetric significance of the HTO form, the health protective assumption generally made in dose estimations, is that all of the releases are HTO, rather than a mixture of HTO and HT.
Some, but not all, of the releases from the nearby Sandia National Laboratories (SNL) appear to be included with the LLNL data. The 1994 Annual Environmental Report (Harrach et al. 1995) notes that the 1994 releases in air effluents from SNL were 3.4 TBq of HTO and 1.5 TBq of HT (90 Ci and 40 Ci) and that those from LLNL were 5.1 TBq with 2.8 TBq as HTO and 2.3 TBq as HT (140 Ci with 76 Ci as HTO and 62 Ci as HT). In 1995 SNL ceased all tritium facility operations. Figure 7-9 in the 1999 Annual Environmental Report (Larson et al. 2000), which shows the correlation between annual median tritium activity in rain and total stack emissions, explicitly includes both emissions from the LLNL site and the SNL site. It would appear that reported releases for LLNL to the atmosphere do not include those of SNL (as illustrated here in Figure 4.2). The more distant ambient monitoring of tritium by LLNL in past years will necessarily be including contributions from SNL. Liquid effluents from SNL have been routed through the LLNL wastewater system since 1980 with combined reporting of total tritium concentrations.
Monitoring of all media in recent years indicates that the levels of tritium (measured as HTO) in other effluents and vegetation in the environment are generally decreasing with time as shown by the 1999 Annual Report (see Figures 4.3, 4.4 and 4.5).
Figure 4.3 shows the general decrease in monthly average concentrations of tritium in the sewer effluent from LLNL site from monthly annual highs in the thousands of Bq/L during the late 1970s down to values in the tens of Bq/L in recent years (from » 100 nCi/L down to » 1 nCi/L ). Similarly, Figure 4.4 indicates the decrease in storm water concentrations from values in the hundreds of Bq/L in the early 1980s down to the order of 10 Bq/L currently (from » 10 nCi/L down to tenths of nCi/L).
Figure 4.5 illustrates the general decrease in concentration of tritium in the tissue water of vegetation with time. From 1971 to 1990 the concentrations were fairly constant at approximately 40 Bq/L (1 nCi/L) at near and intermediate distances from the LLNL site. They have decreased to approximately 4 Bq/L (0.1 nCi/L) from 1990 to 1999.
For tritium analysis, the water samples are first distilled and then counted in a scintillation counter. This procedure should remove the heavier elements such as Cs and Pu, but may not remove all of the tritiated organic compounds. Hence, some of any OBT may be detected in measurements of water samples. For vegetation samples, water has been separated from organic matter by freeze drying so measured tritium will not have included any OBT.
Figure 4.5 Median concentrations of tritium in vegetation sampled on the LLNL site and in the Livermore Valley, 1971 - 1999. For "Far vegetation" for 1998 and 1999, the values are the lowest positive (from Larson et al. 2000; Figure 11-3).
The only measurements that definitely include OBT are those made of tritium in wine. The concentrations of tritium as HTO plus OBT in sampled wines over recent years are shown in Figure 4.6. For the wines made from grapes grown in the Livermore area in 1998, the tritium results ranged from 1.3 to 8.2 Bq/L (35 to 220 pCi/L), for wines from grapes grown in other areas of California, the tritium results ranged from 0.34 to 0.58 Bq/L (9 to 16 pCi/L), and for the European wines, the tritium results ranged from 1.1 to 1.7 Bq/L (30 to 46 pCi/L). These values are well above the minimum detection limit (MDL) of 0.056 Bq/L (1.5 pCi/L) for tritium using the 3He method. In contrast, the MDL for liquid scintillation counting is typically in the range of 4 to 20 Bq/L (100 to 500 pCi/L range).
Figure 4.6 Mean tritium concentrations in retail wines from the Livermore vicinity, from elsewhere in California and from Europe. The tritium concentrations are corrected for decay from the vintage year to the sampling year (from Larson et al. 2000; Figure 11-4).
The Livermore wines do appear to be higher in tritium content than the other California wines tested. The European wines appear to be intermediate. The sampled wines from elsewhere in California had concentrations less than those from Europe. Grapes grown in the Livermore valley do appear to contain some tritium released from LLNL. Overall, in locally produced wines the median concentrations of tritium over the last decade have been less than 10 Bq/L (less than 300 pCi/L).
The total of the measured releases in the 19-year period 1981 to 1999 was approximately 1.2 PBq (30 kCi), so approximately 2.8 PBq (80 kCi) would have been routinely released before this period. Hence, if the period over which releases occurred is taken as 46 years, then the average per year has been approximately 90 TBq, excluding the two incidents noted. The average per year over the last five years has been approximately 7.3 TBq. If these values indicate most of the released tritium, and indicate that the proportion of HTO and HT is approximately the same through the years, then the yearly exposures, very roughly, will have averaged about 12 times those from current releases. In addition, there will have been the exposures to the two major releases in 1965 and 1970.
The concentration of tritium as HTO in waters and vegetation in the public domain in the Livermore Valley currently appears to be generally less than 10 Bq/L (300 pCi/L), with median values in the range of 1 to 6 Bq/L.
For compliance purposes, LLNL is required to estimate doses from airborne tritium releases with the model and computer code (generally referred to as CAP88-PC, and described by Parks 1992) mandated by the Environmental Protection Agency. Three pathways are considered: (1) internal exposures from inhalation of air; (2) ingestion of food and drinking water; and (3) external exposure by irradiation from contaminated ground (zero from tritium). Doses are calculated separately for the Livermore site and Site 300, and also for point source (stack) emissions and diffuse emissions. Doses to three target populations have been emphasized: the dose to the site-wide maximally-exposed individual (who is the hypothetical member of the public at a single residence, school, business, or office who receives the greatest dose resulting from the combination of all radionuclide source emissions from LLNL); the maximum dose to any member of the public (assumed to be at the LLNL fence line); and the collective dose to populations residing within 80 km of the two LLNL sites.
All tritium released is taken to be in the form of HTO and estimates of doses depend on calculating the dispersion of airborne tritium from the site. The environmental-related parameters in the code, such as humidity, and characteristics of the diet, and person-related parameters, such as inhalation and ingestion rates, cannot be altered by the user of the code. An alternative approach (which was taken before the EPA code was required) is to base estimates of dose on actual measurements of contamination in the environment, with assumptions and parameters in the models taken from NRC Regulatory Guide 1.109 (USNRC, 1977). Generally, assumptions are still conservative; for example, in an individual's diet all the vegetables are contaminated with tritium at locally measured values and all meat and milk are from livestock that have been fed with food having the same concentration of tritium. The most recent annual report (Larson et al. 2000) discusses the relative conservatism inherent in them.
For 1999, the dose to the maximally exposed individual, estimated according to the EPA standard method, was 0.9 mSv (0.09 mrem) from point sources and 0.3 mSv (0.03 mrem) from diffuse sources for a total of 1.2 mSv (0.12 mrem). The alternative approach that relates doses more closely to environmental measurements yielded a dose from tritium through all pathways of less than 1 mSv (0.1 mrem). Annual doses to the hypothetical maximally-exposed individual from the ingestion of water, food and wine that was assumed to have the highest concentrations of tritium (as HTO), as measured in 1999 in the respective media, were estimated to be generally less than 1 mSv (0.1 mrem).
LLNL staff has examined the conservatism inherent in the atmospheric dispersion sub-model underlying the CAP88-PC code. Annual-average measured tritium concentrations in 1999 at 12 sampling locations were compared with those predicted by the model with tritium-release data and site-specific meteorological data. The conclusion was the modeling results tended to overestimate the measured tritium concentrations by as much as a factor of seven. There was an indication that there was an increasing overestimation by CAP88-PC with increasing distance so that where the public could be exposed, the overestimation could be greater. The furthest comparison point was about 2 km East-Northeast of the Livermore site. In a recent paper (Parks 1999), additional conservatisms in the model of the CAP-88-PC code, arising from the assumptions made about food consumption are discussed.
As noted above, estimations of doses from tritium that has been released from LLNL, as reported annually, have generally not included any explicit estimation of the contribution from exposures to tritium in the form of OBT. In the most recent annual report (Larson et al. 2000) the potential contribution of OBT is discussed briefly. Doses were estimated based on the measured concentration of water in vegetation collected from LLNL (at the Visitor's Center) and the dosimetric and diet assumptions in the NRC Regulatory Guide 1.109 (USNRC 1997). Without the OBT component, the annual doses from the consumption of vegetables, milk and meat were estimated to be 0.072, 0.063 and 0.073 mSv respectively. For an individual whose diet exclusively comprises vegetables, milk and meat at this level of contamination, the annual dose was estimated to be 0.21 mSv. When the OBT component is included, the doses increase to 0.083, 0.072, and 0.104 mSv (respectively) for an annual dose of 0.26 mSv. Estimations of doses from tritium released from LLNL, as reported annually, have generally not included any explicit estimation of the contribution from exposures to tritium in the form of OBT.
We can compare this with an estimate based on the considerations outlined in Section 2 of this report and applied in Section 3 to the tritium released from the Savannah River Site. We can take the concentration of tritium in the water in foods to be 10 Bq/kg (0.27 nCi/kg), which is the value that in the previous section we concluded was the high end of the range currently for waters and vegetation in the Livermore Valley. The specific activity of tritium will be 90 Bq/kg hydrogen (2.4 nCi/kg). If we assume that all the tritium in the moisture in a reference daily food intake of 1.6 kg has this specific activity and that the specific activity of the tritium is 1.2 times this - i.e., 108 Bq/kg hydrogen (2.9 nCi/kg), then, with a total daily hydrogen intake of 0.14 kg and 85% moisture in the food, the annual dose from food would be 0.11 mSv (11 mrem) with 36% coming from the OBT intake. This estimate of annual dose is similar to that in the above paragraph. Both approaches have entailed inherently conservative but different assumptions.
The ambient concentration of tritium in waters and vegetation in earlier years was an order of magnitude greater than currently (see Figures 4.3, 4.4, 4.5). It follows that annual doses from OBT in foods, estimated with the same assumptions as above could have been of the order of 1 mSv (0.1 mrem).
Clearly, based on the measurements of tritium in effluents and in environmental media and with the modeling and dosimetric assumptions detailed in the LLNL Annual Environmental Reports, the annual doses to the members of the local Livermore valley population from tritium are very low - less than 1 mSv (0.1 mrem). There are some conditions on this statement:
- There are no significant other sources of airborne tritium -- of a form that would not be collected by molecular sieve or silica gel samplers-- that could give a significant dose by inhalation or that could contaminate foodstuffs without appearing as HTO. That there would be such sources seems very unlikely.
- Any OBT formed in foodstuffs from the HTO and HT emissions is insufficient to cause the estimated doses from tritium to be underestimated, given the reasonably conservative assumptions that are made in the dose estimates. Although it is not known what the levels of OBT actually have been or are in foodstuffs there is good evidence that the ratio of specific activities of tritium in the water and organic components of food does not differ substantially from unity. Given the low ambient concentrations of tritium in moisture, the ratio of tritium as OBT to tritium as HTO would have to be more than 200 times greater than unity for the doses to approach regulatory limits.
As noted above, annual airborne emissions in the past, as reported, have averaged about 12 times the current five-year average value. If airborne emissions are taken as the dominant source of doses from tritium to the local population, then it would follow that average annual doses over the past years could have been greater in the same proportion. The conditions on this statement are as follows:
- The reported values for measured emissions in the past, when releases were much higher, are accurate. Early monitoring may not have been carried out with the same QA as is currently the case.
- The releases from Sandia National Laboratories, adjacent to the LLNL site, are accounted for via LLNL environmental measurements.
- The HT and HTO ratio in stack effluents has not changed greatly.
- The tritium in liquid releases in the early years did not result in large off-site doses.
- The unplanned releases of HT in 1965 and 1970 (350 kCi and 289 kCi, respectively) will have contributed to the doses for those years. For the 1970 release, Myers et al. (1973) estimated that the "maximum credible dose" to a child was 0.7 mSv.
There is a further uncertainty in that there have been only few validations of the atmospheric dispersion model used for estimating concentration of tritium in air more than 2 km from the site. The observation that the measured air concentrations at Zone 7 were below those predicted by the CAP88-PC modeling results during 1999 (LLNL 1999b) does not by itself provide confidence that the model will be similarly conservative or even consistent at all of the locations where the model is used to calculate off-site tritium doses.
In summary, the uncertainties in estimating the doses from tritium to the local population over the past years mainly rest with the accuracy with which early releases are known, and the actual OBT levels that might have been attained in vegetation. Measurements of tritium in sampled environmental media and biota provide assurance about the levels of contamination actually present but it would be helpful to tie this in with validation of the atmospheric dispersion modeling.
There is no evidence of any food item constituting a particular hazard because of a high OBT content relative to its HTO content. For precise estimates of dose, actual separate measurements of OBT and HTO would be needed. However, the extensive observations of the OBT/HTO specific activity ratios now reported in the general literature provide a good guide to values that can be taken for such ratios in environments chronically contaminated with tritium from HTO and HT emissions, as is the case in the Livermore Valley.
The implication for LLNL is that some measurements of OBT and HTO in locally produced foods would indicate whether conditions or biota were sufficiently different for the ratios of specific activities of tritium as OBT and HTO to be significantly greater than those observed elsewhere.
5 In Ci (approximately): 1999 (210, 70); 1998 (80, 25); 1997 (270,32); 1996 (180,32); 1995 (62,30)