KARST AQUIFER TEST REPORTS
LEMON LANE LANDFILL

1.0 Introduction

Four tests were conducted. The testing period began on October 31, 2001 and was completed on November 23, 2001. During this time there were no major storm events. Figure 1 is a plot of measurable rain fall and mean temperatures in the Bloomington area during the test period. The test period was bounded by significant rain on October 24, 2001 and November 25, 2001. Temperatures were typically mild for November.

The first test was a dye injection. The dye injection occurred on October 31, 2001 with follow up sampling conducted at various locations from two days to several weeks. On November 7, 2001 a pump test was conducted at borehole MW21. This well is located on the east side of the landfill. This test lasted until the afternoon of November 9, 2001. On November 13, 2001, a second pump test was conducted at borehole MW16. This test was concluded on November 19, 2001. Finally, an injection test to Sargent's Pond was conducted on November 20, 2001. Monitoring for the injection test ceased on November 23, 2001.

The following sections describe the test results.

2.0 Dye Testing

2.1 Introduction

For each of the injections, 300 grams of dye were mixed with a gallon of water. For wells NN700 and SP 1, this water was then injected using a peristaltic pump with discharge tubing placed at the 795 elevation. For LF6-8 the dye/water mix was just poured down the well casing. All initial dye injections were followed by 110 gallons of flushing water. The bottom of the LF6-8 casing is sumped in approximately 3 feet of clay, which holds about 8 gallons. So residual dye remained in the hole after initial flushing. In addition to the injections on October 31, 2001, water was also flushed into LF6 on two additional occasions and into NN700 and SP1 on one separate occasion. These additional flushes of water were done in concert with pump test conducted in November 2001, and were intended to flush residual dye out of the well bore. The water flush results will be discussed separately in the report on the pump tests.

Wells around Lemon Lane and four springs (Illinois Central, Quarry B., Slaughterhouse and IC Weir) were monitored for the dyes injected on October 31 by sampling at various periodicity. The test plan, Attachment 1, was essentially completed.

The goals of the dye test were basically threefold. First, to determine if any of the wells surrounding the site were on or near a major conduit which would integrate all three dyes into one flow path towards the IC Spring. Second, to determine travel times from various areas of the site to the IC Spring. Third, to determine if Quarry B would receive dye while flow was stopped to the swallow hole. The Quarry B goal was done in concert with the EPA who is trying to determine if underflow occurs between IC and Quarry B Springs.

All the dye results have been summarized on tables by sampling locations. These tables are presented as Attachment 2. The attachment includes dye sampling conducted in concert with the pump testing performed at later dates and also discussed in Section 3. The initial dye results from the October 31 injections will be discussed on a dye by dye basis in this report. First, background results are discussed in Section 2.1.1.

2.1. Background Fluorescence Samples

Attachment 2 includes the samples, locations and results taken for background fluorescence prior to the dye injection. Not every station sampled had background samples taken, as some stations, e.g. Valhalla "A" wells, were added to the sampling plan during the test.

There are several sources for background fluorescence. Natural occurring organic compounds such as certain algae, chlorophylls, and manure leachates can fluoresce in the blue-green range near Fluorescein. Anthropogenic sources include radiator coolant (Fluorescein), brake and transmission fluid (Rhodamine which would be similar to Phoxine B), and septic discharge containing laundry detergents (blue-green opticalbrighteners) and shampoos (green and red dyes). However, previous dye trace injections at Lemon Lane are by far the greatest source of background fluorescence. These include.

• 1989 Direct Yellow in wells MWls and MWld
• 1989 Optical Brightener in MW7
• 1989 Fluorescein in MW10
• 1990 Rhodamine WT in MW10, MWls, MW7
• 1992 Fluorescein, Rhodamine WT, Eosine, Optical Brighteners, and Direct Yellow in five sinkholes around Lemon Lane
• 1996 Fluorescein in swallow-holes in southwest corner of landfill
• 1996 Rhodamine WT in North Sink (900' north of the site)

It has been discovered recently that Rhodamine WT, once in the environment, breaks down into compounds that fluoresce anywhere from 520 nanometers to 578 nanometers.

Illinois Central Spring had background fluorescence for both Fluorescein and Rhodamine WT. If the samples between injection and first detection on the breakthrough curve are counted as background, then background Fluorescein ranged from 0.11 to 0.13 ppb.

Quarry B spring also had minor amounts (less than 0.1 ppb) of Fluorescein in the background samples as did the Swallow-hole seep.

Minor amounts of Fluorescein were seen in 00-370 and 00-300. A Phloxine-like fluorescence was seen in 00-587 on a lOxlO scan, but does not appear on any of the 5x5 scans during the test. NN-625 has a small amount of degraded Rhodamine WT and a 0.7 ppb background Fluorescein, but neither one shows up again during the test. An Eosine-like fluorescence appears in a MW6 background at lOx10 scan, but does not reappear in any of the 5x5 scans during the test.

MW15 was the only Lemon Lane well to show Fluorescein background, besides MW1O. No Fluorescein appeared in the well during the test. MW10 had Fluorescein injected in it in 1989 and showed a 37 ppb residual level in its background sample. MW1O also had high amounts of Rhodamine WT residual. Because of these residuals, MW10 was not sampled during the test. MW16, 18, and 19 had background samples clear of fluorescence. The other wells near MW10 (MW20, MW21, MW4i, MW4s, and MW17) showed varying amounts of degraded Rhodamine WT, in background, as well as during the test as did MW2. These high degraded Rhodamine WT residuals were probably a contributing factor to the lack of detection of Phloxine B during the test. These high backgrounds also make any isolated dye detection in the absence of a breakthrough curve during the test suspect.

2.2 Fluorescein

2.2.1 Well Samples

It should be understood that the method of sampling wells might lead to some confusion about the results. All of the wells but one were generally sampled by hand using a dedicated bailer in each well. The bailer would be pulled up and emptied into a sample jar. Then the bailer would be dropped back down the well and suspended until the next sample time. Since bailers fill when dropped down a well bore, the sample time listed for any well sample sampled in this manner can be misleading. The samplers were generally agitated up and down in the well bore before pulling the bailer up at the sample time. This would mix the water in the bailer. So it would be most correct to consider any sample taken this way to be a mixture of water that was in the hole at the time it was placed down the hole (the previous sample time) and the present sample time water. In other words a sample listed as taken at 12:00 may actually be a mix of conditions at the well the previous sampling time and the current sample time.

Another important fact about this method is that the well is not purged at all with this method. This implies that the well would have to be very close to or on a conduit to receive a significant hit of dye. The exception to this sampling scheme was MW21. This well was purged and sampled with a submersible pump each time.

These Fluorescein results are significant in that Viacom had always assumed that the water in the epikarst at Lemon Lane would flow down dip (to the southwest), then descend to the phreatic zone, and flow laterally to the spring in a conduit that would generally following the non-storm potentiometric surface. However, this hypothesis was based on examples given by Palmer¹ (1999) for other karst aquifers. These dye results do prove that at least some of the water/dye mix injected at LF6-8 (an epikarst well) did flow to the southwest into Valhalla cemetery. They also suggest that this flow occurred in the epikarst to a point beyond NN300, then some of the dye descended to the phreatic zone and flowed back toward the site (this would match the direction of the low flow potentiometric surface in the area of the site). This would explain the appearance of dye in well 00370 before well NN300.

Note that wells NN300A and 00300A are shallow wells that do not extend all the way to the 795-800 feet amsl horizon while most of the other wells (such as 00370 and NN300) do. These shallow wells were not sampled at the same frequency as the deeper wells, nor were background samples taken. The first samples taken from these wells were the November 1 11:30 samples. The dye was already present in high concentrations, so minimum travel times cannot be ascertained. These shallow wells have by far the highest concentrations of Fluorescein found in any well or spring sample, which implies they are nearest a primary pathway. It also implies the primary pathway is above the 795-800 horizon.

Well 4S is the only non-Valhalla well which received an apparent large hit of Fluorescein. Since only one sample at this location was positive for the dye, there is some question about the validity of this result. It should be noted that 22 kilograms of Fluorescein was injected in well MW10 in 1989 and 37 ppb residual concentration was noted in that well in the background sample. All other locations that tested positive for Fluorescein had more than one sample from that location containing the dye. The timing of the sample also causes some concern for its validity. The sample on Table 1 from well 4S that is listed as collected at October 3 l at 17:35 may in fact have contained mostly water that was in the well column at 13:40 on October 31 when the bailer was last dropped down the well column. Since well LF6 was injected at 12:00 on October 31, it seems unlikely that Fluorescein would have made it to well 4S within 1 hour and 40 minutes. Therefore, the detection of Fluorescein in this well is more questionable than the other wells.

On the other hand, Mike McCann took this sample and recalls this particular sample being tinted green in the bailer. This would indicate that it is not a lab analysis error, or a labeling problem. Another indication that this is a real detection is the detection of Fluorescein in MW21 once pumping of this well started on November 7, 2001. Low levels of Fluorescein show up in this well within a half hour of starting the pump. This would indicate some of the dye residual was near the well. Although note that this residual could be from the 1989 injection at MW 10. Well 21 and 4S are within 10 feet of each other. If this is a valid detection of the Fluorescein injected in LF6-8 in well 4S, it would indicate that a very fast connection (a conduit) exist between LF6-8 and 4S.

2.2.9 Spring Samples

Although small background levels were detected in the Swallow-hole Seep and at Quarry B. a distinct breakthrough curved for Fluorescein was seen only at IC Spring. See Figure 3. This shows the Fluorescein concentrations over time at IC Spring for the dye test period. Note there is a "saw toothed" peak period between November 1 14:00 end November 2 14:00 followed by a single large maximum at November 2 20:00. This would indicate that there are at least two flow paths from the injection point to the spring. The first arrival at IC spring occurred at 09:00 on November 1, 2001. This is a travel time of 21 hours. The second flow path has a first arrival time of approximately 52 hours after injection (this arrival time is most likely skewed later than the actual time since the first dye arrival is so significant) and there is peak overlap.

To further evaluate the dye recovery at the IC Spring, a computer program developed by EPA for quantitative dye recovery analysis in karst terrane was run on these results. Table 2 summarizes the pertinent features of this dye injection from the computer model. As can be seen from the table, there is much more dye mass calculated to be recovered at the spring than was actually injected during the test. Usually, much less dye is recovered than was actually injected due to sorption and dead space/dispersion losses along the aquifer.

This indicates a significant error in one or more of the parameters used to calculate the mass recovery (water flow and dye concentrations at the spring). Since water flow is calculated from sump fill rates (which should be accurate with design and/or as-built drawings), and seems to not be significantly off, based on time of year and proximity of rain events, the evidence points to dye concentrations as the most likely error. Unless this is resolved, all users of the data should be cautioned that the dye concentration data should be considered semi-quantitative. Some likely causes of quantitation error include:

• The unknown impact of a mixture of dyes on machine calibration. Note that the calibration curves were run on single dye mixtures. While in reality three dyes were injected and all three could have arrived at the spring at similar times resulting in a 3 component mix. The impact on quantitation for one dye of a 3 component mix is not known.
• The calibrations were done with distilled water. Some unknown component in the actual spring water could be causing an error. Background samples at the spring did indicate some interfering amount of Fluorescein at levels from 2-24 ppb prior to injection. This high background will cause some error, but not the magnitude of error apparent in this data.
• Most of the 3x3 scans conducted on samples from IC Spring had results which were above the calibrated range of the instrument. For example, the highest 3x3 calibration standard run had an intensity of 160 IU. Out of 40 3x3 scans run, only 3 had an intensity lower than 160. Typically, samples with an intensity higher than the highest calibration standard should be diluted to an intensity value below the highest calibration standard to ensure accurate results. By contrast, only 1 out of almost 100 SxS scans were reported at an intensity above the calibrated range. This implies the SxS scan data is more accurate then the 3x3 data. The 3x3 scans were used almost exclusively during the heaviest dye arrival period at the spring. A large error in those samples could cause a large overall error in recovered dye mass.

The first arrival time of 21 hours after injection compares to a PCB travel time of 14 hours at the actual flow rates experience during the test period. This means that the PCB pulse during a small storm event would be expected to beat the dye to the spring by 7 hours. This difference could be due to some allowance for travel in the epikarst. It is not known if the PCB travel time correlation with flow developed from storm data is applicable to the dye travel time found during this non-storm condition. If it is directly applicable, this travel time would indicate that the PCBs which first arrive at the spring during storms take a shorter route than the dye injected at LF6-8. A dye injection in LF6-8 concurrent with a storm event would clarify this.

There was one very low level Fluorescein detection at Quarry B spring on November 1 at 01:00. The level detected was lower then some background samples at the spring and also the detection occurred prior to detections at IC Spring. These two facts decrease the probability that this is a valid detection of the Fluorescein injected during this dye test.

Fluorescein was not detected at Slaughterhouse Spring. This spring was sampled for almost two days after injection.

2.3 Eosine Detections

2.3.1 Wells

Eosine was detected in the following wells around Lemon Lane:

MW 15,16, 18,19,20,21 and MW2 on the east side of Lemon Lane and NN 625, 00587, 00370, NN300 and 00387 in Valhalla. These results are summarized in Table 3. All of these detections are at much lower levels than found for the Fluorescein.

By far the highest level detected on the east side of Lemon Lane was found in MW20 and MW 21 and in Valhalla in 00587. The other levels of detection in the east side wells are low. This causes some question about the validity of the detections on the east side in wells other than 21 and 20. However. the detections on the east side of the landfill make sense with the injection time. There was also no cosine detected in any pre-injection background samples for these wells. Therefore in the aggregate, we should assume these are valid detections.

The first detection of the cosine was in 00370 on October 31 at 14:26. This was followed by detections at MW16 (17:10), MW21 (19:00), MW2 (20:53), MW18+19 (21:50) and MW20 on November 1 at 01:28. This would indicate that water flowing near NN700 generally flows in an eastnortheast direction. With generally the higher levels on the east side found in MW21/20, this indicates that these wells are nearer a collection point for this water. Figure 4 depicts an interpretation of these results.

2.3.2 Springs

Eosine was not detected at any of the springs after the initial injection. The two most likely explanations for this are that either the cosine arrival at the IC Spring was masked by the much larger Fluorescein detections at IC Spring (implying that its travel time is slower) or that the cosine was diluted beyond detection limits. IC Spring was intensely monitored for an extended period of time after injection. Slaughterhouse Spring was monitored sporadically for up to 1 week after injection.

2.4 Phloxine

Phloxine was not detected unambiguously in any well or spring until pumping of MW2 1 began on November 7, 2001. There are several possible reasons for this. First, it is not known if enough phloxine was injected. Its sorption properties may be such that it requires a much more massive injection than we performed to be detectable in our system. Second, it may have been at the IC spring at the same time the more massive amounts of Fluorescein were present and the Fluorescein may have masked the phloxine. Third, the dye could have gone to an un-monitored location, or arrived at a monitored location after monitoring ceased.

After pumping MW21 for 30 minutes, phloxine seemed to be present in all pumped samples from this well at low levels. This indicates that the phloxine path was near this well, but not close enough to be seen soon after the initial injection without more extensive pumping.

2.5 PCB Results in Spring Waters

PCB sampling was conducted at IC and Quarry B Springs. The results of those samples are summarized in Table 4. All but one of the original PCB results reported for the IC Spring were rejected by Viacom as biased low by the laboratory.

Some archived samples for both springs were submitted to the lab in an attempt to recover data for the period in question. Viacom suspects that the archived sample results are biased low because they were not properly preserved and seriously exceeded hold times. Even with an assumed low bias, the archived sample results show that the original results reported for all but one of the IC Spring samples were biased very low.

Viacom's investigation showed that all the biased low results were prepped in one batch by the same analyst. This indicates an error at the prep stage. The Quarry B samples were prepped in a separate batch by a different analyst and there is no indication that the original reported results for Quarry B Spring are biased.

The purpose of sampling at IC Spring for PCBs during the dye test was to determine if a slug of PCBs would arrive with the Fluorescein injected at LF6. Of the 11 samples taken at the spring during the dye test, only the last sample result is considered valid. This one sample result was taken on November 2 at 05:00. This was in the middle of the breakthrough curve period for Fluorescein at the spring. The PCB result for this sample (11 ppb) is higher than expected based on pre-remedy historic trends but not by a wide margin and is somewhat consistent with data collected post remedy.

The purpose of collecting PCB data at the Quarry Springs was to determine if PCBs would fall off after water was prevented from entering the swallow hole. The data does not show a definite trend.

2.6 Conclusions for the Dye Test

Some of the Fluorescein flushed from LF6 traveled to Valhalla Cemetery in vadose passages.
• Some of the Fluorescein did descend to the phreatic zone near well 00370.
• The Fluorescein response at IC Spring indicates there may be more than pathway from LF6 to the spring.
• No dye was detected at Slaughterhouse Spring. This spring was sampled sporadically for up to 1 week after injection. This indicates that this spring does not receive site related waters quickly and is consistent with previous dye testing that showed this site to most likely be a minor discharge area for site waters.
• There was no breakthrough curve for Fluorescein at Quarry B Spring or the weir seep. This indicates that these springs are not carrying IC Spring underflow waters.
• Eosine appeared to travel to the east-northeast.
• Eosine was not detected at any spring. This implies either that the Fluorescein path to IC Spring is quicker than the cosine path or that the cosine was diluted below detection at the springs.
• Phloxine was not detected. This implies either that the Fluorescein path to IC Spring is quicker than the phloxine path or that the phloxine was diluted below detection at the springs.

3.0 Pump Testing

MW2 1 and MW 16 are wells on the east side of the Lemon Lane Landfill. These wells were installed as part of the karst conduit system investigations in 2001. As part of these investigations, these wells were pumped for extended periods of time in November 2001. This report summarizes the data and suggests follow up investigations based on this data. The test plan for pumping is included in Attachment 3.

During these pump tests, water was also flushed down the following wells:

These water flushes were done to provide a pulse of dyes at these locations while pumping. Dyes had previously been injected at these wells on October 31, 2001. It was suspected that by merely flushing water at these wells, the residual dye at these locations would be flushed into the karst system and provide additional information about the radius of influence/capture potential of each pumping well. This report also discusses the recovery of dye at the pumping wells and IC spring aRer each of the water flushes. The following sections discuss each test.

3.1 MW2 1 Pump Test

The pump test at this well began on November 7, 2001 at 10:00 am. The well was pumped with a 4 inch submersible pump that was lowered to the 800 amsl elevation. The pumping rate was set at 17 8pm and maintained for the duration of the test. The pump was shut off on November 9 at 14:00 for a total of 52 hours of pumping. A total of approximately 65,000 gallons of water was pumped from the well. The water was treated on-site and released to the lined retention pond in the southwest corner of the site. The retention pond outlet was blocked so that none of the treated water could escape.

The original plan was to pump for at least three days. The pumping was halted after 52 hours for two reasons. First, the preliminary report on the PCB samples from the pumping well and the spring were not encouraging. Second, the post water treatment PCB levels were slightly above the discharge limit indicating that a shutdown was necessary to correct the treatment process.

Wells in the vicinity of the site were monitored for water level before, during and after pumping. Flow and Conductivity at IC Spring was continuously monitored by IDEM at their treatment facility

In addition to sampling the well for PCBs during pumping, the well was also sampled for dyes. This was done because of the dye injections done on October 31, 2001. It was thought that residual levels of dyes from those injections may still be in the aquifer and the arrival times of the dyes at the well may provide insight on the closeness of the well to the dye travel paths. Additionally, well LF6 was flushed with 300 gallons water on November 8 at 15:30. The injected water was water previously pumped and treated from MW2 1. Dye sampling continued at the well to determine if water flushed on November 8 would result in a dye hit at the pumping well.

A major goal of the pump test was to determine if PCBs at the IC Spring would be impacted. It was hoped that eventually, high levels of PCBs would be withdrawn from the pumping well and this would result in a substantial drop in PCBs at the spring. Samples were taken at the spring periodically during the duration of pumping and analyzed for dye and PCBs.

The results of the MW21 pump test are summarized in Sections 3.1.1 - 3.1.6 below

3.1.1 PCB Results at the Pumping Well

A summary of field parameters and PCB sample results for the MW21 well is shown in Table 5. The PCBs at the pumping well started out and finish in the 5-7 ppb range. This is not seen as a major change. Therefore, it can be concluded that the MW21 well is not on or hydraulically well connected/near a major conduit that carries a large amount of PCBs to the IC spring. The 5-7 ppb range compares to PCB levels at the IC spring which are typically 10-20 ppb under these flow conditions.

It is not clear why this well would produce so much less PCBs upon pumping than did MW4I. This well is only 10 feet from 4I and the pump was placed at about the same elevation in this well as when 4I was pumped. It was expected that it would produce a similar level of PCBs. MW4I produced high levels of PCBs during two separate pump tests in the summer of 2001. However, 4I was pumped for a short period of time on November 6, 2001 just prior to starting the MW21 pump test and it showed only low levels of PCBs (1-2 ppb). It should be noted that well MW4I was modified by removing some bentonite at the bottom of the well between the summer pump test and the mini-pump test performed on November 6. This modification could have changed the zone that supplies most of the water to the well upon pumping.

Because there were some PCB samples analyzed at Heritage Labs that were found to be biased low, it is recommended that additional samples of well 4I be taken at the earliest opportunity to determine if this well can once again produce high levels of PCBs. These samples should be taken with some limited pumping. To determine if the hydraulics at the well have change as a result of removing bentonite, the drawdown should be monitored during the short pump test for comparison to the summer data.

3.1.2 Aquifer and Well Hydraulics

Table 6 provides a complete listing of the continuous monitoring well level data. The pumping well drawdown and elevations are listed on Table 5 and shown in Figure 5. Elevations for the other wells are shown on Figure 6. The drawdown and field conductivity for the pumping well show that hydraulic equilibrium was achieved. The final elevation reached at the pumping well was 814.4 feet amsl. The total drawdown for the pumping well was 1.88 feet. This compares with over 5 feet of drawdown in well 41 when it was pumped at a lower rate in the summer of 2001. Even though the overall aquifer conditions were drier in the summer, this large of a difference in drawdown most likely indicates that this well is much better connected to a water source than well 41. Well 41 is a smaller well bore than MW21 and this smaller bore could partially explain why so much more drawdown was achieved in MW4I. But the difference is so dramatic other factors are likely at play.

A more thorough analysis of the drawdowns for the pumping wells and monitoring wells during both this and the MW16 pump test is included in section 3.3. Figure 7 and Table 7 shows the flow and conductivity at IC Spring during the pump test. The flow at IC spring dropped almost immediately upon the commencement of pumping. While the conductivity continued a fairly normal receding limb/base flow rise. Flow also responded almost immediately to the cessation of pumping. Conductivity appeared to stabilize and then slightly fall the next day after pumping ended. Whether or not this is caused by the end of pumping is confounded by a small amount of rain (.02 inches) that fell on November 8.

The rapid response of both the spring flow and remote well levels indicates that the pumping well is very well connected to the aquifer, and that all the wells and springs are well interconnected. It should be noted that most of the monitored wells were drilled to at least the same elevation as the pumping well. But one well, 4S is drilled several feet shallower, this well also responded similar to the deeper wells. No significantly shallower or deeper wells were monitored for water level during this pumping period to determine if they would be similarly impacted.

3.1.3 PCBs at IC Spring

The PCB response at IC Spring to pumping at MW2 1 is shown on Table 7 and Figure 8. As can be seen, PCBs were at about 12 ppb prior to pumping and rose to a range of 18 to 24 ppb within 7 hours of pumping. The PCBs decreased slightly to a range of 17 to 21 ppb range after 28 hours of pumping. The PCBs dropped to the 10 ppb range within 8 hours after the pumping ceased. They stayed in this range for about 28 hours then rose to the 12 to 18 ppb range.

One of the main goals of the pump test was to determine if significant levels of PCBs would be drawn from this well while pumping and if the PCBs at the IC spring would fall during the pumping period. This did not happen. The spring PCB levels increased during the pumping period, drifted slightly lower while pumping and then appear to have fallen after turning the pump off. This indicates that if there is a conduit system near MW21, it may be providing relatively clean water to the spring.

The PCB results at the spring must be observed with some caution. It appears that some results may be affected by the amount of time the bottle stayed in the autosampler and the amount of heat built up in the sampler during that time. Evidence for this impact was observed during periods of high temperatures most recently during the July/August pump testing done at well 41. During this test period, most samples that stayed in the sampler for more than 24 hours have lower PCB results than those that were in the sampler for less time. For this test period, the samples from November 9 20:00 to November 11 16:00 were in the sampler for more than 24 hours. The results for those samples may be biased low by what appears to be in sampler volatilization losses. Therefore, the conclusion that PCBs fell at the spring after turning the pump off on November 9 is not certain.

3.1.4 Dye Response at the Pumping Well

Dye was sampled at the pumping well to determine if any residual from the dyes previously injected in the aquifer on October 31 would be drawn to the well while pumping. If and when this occurred may provide an indication of how far the well is from a flow path carrying the dyes. The dye sample results are shown on Table 8 and Figure 9.

As can be seen on the table, only a degraded rhodamine (DRWT possibly from a past injection) was present in the well as pumping began. But within one half hour of pumping both Fluorescein and Phloxine became detectable. All three dyes were routinely present in all pumping samples thereafter. A large Fluorescein peak occurred at November 9 at 16:00. This is thought to be a result of the flushing of LF6 on November 8 at 15:30.

• This well did receive an cosine breakthrough curve during the dye test without pumping.
• This indicates it is near a path that carried cosine from the NN700 area (the southwest corner of the site).
• This well did not have Fluorescein or Phloxine hits during the October 31 dye test but did receive these two dyes within 30 minutes of the start of pumping the well on November 7. This indicates that at least some portion of these dyes flowed near the well.
• The well received a hit of Fluorescein after flushing LF6 on November 8. This indicates that at least some of the Fluorescein flushed from LF6 does get captured by this well when pumping. However, the IC Spring also received a large hit of Fluorescein from the November 8 flush indicating that pumping this well cannot capture all of what is flushed from the LF6 area. Relatively speaking, the amount of Fluorescein recovered at the spring after the flush was much larger than that recovered at this well. This indicates there are two flow paths for waters flushed from the LF6 area and that this well is near the smaller of the two. This further indicates that this well is not an efficient pumping location to capture materials flushed from the epikarst at LF6. This is supported by the low level of PCBs recovered in the well since the epikarst at LF6 has been shown to contain high levels of PCBs.

3.1.5 Dye Response at the IC Spring

Dye samples at IC Spring were taken for some periods during the pump test period. See Table 7 for a complete listing of the results. Only Fluorescein was reliably and consistently detected at the spring. The analysis of the Fluorescein response to pumping at the spring is confounded by two factors. First, the Fluorescein concentrations at the spring were receding from the October 3l dye injection. Second, samples for dye analysis were not taken from the period 6 hours prior to the start of pumping until about 20 hours after pumping began.

Figure l0 is a plot of Fluorescein concentrations at the spring during the MW21 pump test. While Fluorescein concentrations during pumping are at no time higher than pre-pumping levels, they may be higher than would be expected had the Fluorescein continued to decline at the same rate occurring just prior to turning the pump on.

There is a large amount of dye indicated by a breakthrough curve for Fluorescein that begins on November l 0 at 08:00 (after pumping stopped the day before). It is believed that this was a result of the flushing of water at LF6 that occurred on November 8 at l 5:30. Note that there is only a single peak to this curve.

3.l.6 Conclusions and Recommendations for the MW2l Test

Pumping MW2 l did not lower the PCBs at IC Spring after 52 hours. It also appears that pumping this well will not capture most of the materials that are flushed from the epikarst near LF6.

Hydraulic indicators such as low flow water levels and quick equilibration during pumping (along with relatively low drawdown per gallon/minute pumped) show that this well is near or on a major conduit. This conduit most likely drains the 795-800 elevation zone around the site and possibly all zones in the northern portion of the site.

These results along with the previous dye test indicate that there are at least two major flow networks feeding the spring from the site area. It appears that most of the PCBs released from the southern epikarst region do not use the flow network associated with this well to reach the spring. These results along with the previous dye testing suggest that either:

• The flow network carrying the PCBs from the southern epikarst is at a different elevation. This is because all the wells around the site that were monitored at the pumping well elevation responded in a manner which indicates they were connected to the pumping well.
• Or the flow network is at a similar elevation, but much of the transport occurs in the epikarst and the connection between zones occurs so far from the pumping well that it was not affected by pumping this location.

It is not clear what happened to the high PCB levels previously found in well 4I. This well should be resampled to determine if the previous high levels were just a random flush of a small pocket rather than a long term issue.

3.2 MW16 Pump Test

The pump test at this well began at November 13, 2001 at 12:00. The well was pumped with a 4 inch submersible pump that was lowered to the 800 amsl elevation. The pumping rate was set at about 17 8pm and maintained for the duration of the test. The pump was shut off on November 19 at 08:30 after a total of 140 hours. A total of approximately 143,000 gallons of water was pumped from the well. The water was treated on-site and released to the lined retention pond in the southwest corner of the site. The retention pond outlet was blocked so that none of the treated water could escape.

The test was continued for significantly more than three days because the MW2 1 test was cut short and it was desired to make sure that adequate time was allowed for full hydraulic and chemical equilibrium to be established.

Wells in the vicinity of the site were monitored for water level before, during and after pumping. Flow and Conductivity at IC Spring was continuously monitored by IDEM at their treatment facility.

In addition to sampling the well for PCBs during pumping, the well was also sampled for dyes. This was done because of the dye injections done on October 31, 2001. It was thought that residual levels of dyes from those injections may still be in the aquifer and the arrival times of the dyes at the well may provide insight on the closeness of the well to the dye travel paths. Additionally, well LF6 was flushed with 300 gallons water on November 16 at 15:30. This water was well water that had been processed thru the treatment trailer. Dye sampling continued at the well to determine if water flushed on November 16 would result in a dye hit at the well.

A major goal of the pump test was to determine if PCBs at the IC Spring would be impacted. It was hoped that eventually, high levels of PCBs would be withdrawn from the pumping well and this would result in a substantial drop in PCBs at the spring. Samples were taken at the spring hourly or every 4 hours during the duration of pumping and analyzed for dye and PCBs.

The results ofthe MW16 pump test are summarized in sections 3.2.1-3.2.6.

3.2.1 PCB Results at the Pumping Well

A summary of field parameters and PCB sample results for the pumping well is shown in Table 9. The PCBs at the pumping well started out at about 8 ppb and slowly declined to a minimum of .6 ppb and then stabilized in the 2.5-3 ppb range. These results are generally consistent with the short term pump test conducted in the spring of 2001 at this well.

It can be concluded that the MW16 well is not on or hydraulically well connected near a major conduit that carries a large amount of PCBs to the IC spring. The 2-3 ppb range compares to the 5-7 ppb range withdrawn from MW21 while pumping and with PCB levels at the IC spring which are typically 10-20 ppb under these flow conditions. Based on these results, this well is not as well positioned as MW21 for PCB recovery.

3.2.2 Aquifer and Well Hydraulics

The pumping well drawdown and elevations are shown in Figure 11 and Table 9. Elevations for the other monitored wells are shown on Figure 12 and Table 10. The drawdown and field conductivity for the pumping well show that hydraulic equilibrium was achieved. Note that the apparent large decline in conductivities at 16:00 on November 15 is believed to be an instrument problem. All conductivities after this time are considered biased low.

The final elevation reached at the pumping well was 814.38 feet amsl. The total maximum drawdown for the pumping well was 1.94 feet. This compares with over 5 feet of drawdown in well 41 when it was pumped at a lower rate in the summer of 2001 and with 1.88 feet seen when pumping MW21. Even though the overall aquifer conditions were drier in the summer, this large of a difference in drawdown most likely indicates that this well is much better connected to a water source than well 4I and similarly connected to the aquifer as MW21.

A more thorough analysis of the drawdowns for the pumping wells and monitoring wells during both this and the MW21 pump test is included in section 3.3.

Figure 13 shows the flow and conductivity at IC Spring during the pump test. The flow at IC spring dropped almost immediately upon the commencement of pumping. While the conductivity continued a fairly normal receding limb/base flow rise. Flow also responded almost immediately at the cessation of pumping. Conductivity appeared to stabilize and then fall the next day after pumping ended. This is the same trend as seen at the end of the MW21 pump test. The interpretation of a conductivity fall the day after pumping ceased is somewhat confounded by a small rain event that occurred on November 19 (about .2 inches of rain fell between 07:00 and 13:45).

The rapid response of both the spring flow and remote well levels indicates that the pumping well is very well connected to the aquifer, and that all the monitored wells and IC Spring are well interconnected. It should be noted that most of the monitored wells were drilled to at least the same elevation as the pumping well. But one well, 4S is drilled several feet shallower, this well also responded similar to the deeper wells. No significantly shallower or deeper wells were monitored for water level during this pumping period to determine if they would be similarly impacted.

3.2.3 PCBs at IC Spring

The PCB response at IC Spring while pumping at MW16 is shown on Table 11 and Figure 14. As can be seen, PCBs were in the 12-14 ppb range prior to pumping and rose to about 20 ppb within 5 hours of pumping. The PCBs remained at these higher levels until about 26 hours after pumping began when they dropped to the 7-1 1 ppb range for about 90 hours. PCBs then rose back to the 10-16 ppb range. After 72 hours of pumping the PCBs again dropped to the 10 ppb level for approximately 16 hours and then they fluctuated between 10 and 16 ppb for the next 24 hours before rising back to the 1~20 ppb range for the final twelve hours of pumping.

It is not clear why there would be so much apparent cycling of PCB levels during this test. There was no significant rainfall during the test and hydraulic parameters at the pumping well and observation wells appeared stable. The only perturbations introduced to the system was a series of three water flushes. The first flush occurred in SPI on November 14 at 15:30. The second occurred on November 15 at 1 1:30 at NN700 and the third occurred on November 16 at 15:30 at LF6. The NN700 flush was 100 gallons while each of the others was comprised of 300 gallons of treated well water. It is believed that these flushes of clean water were too small to explain the relatively large changes in PCB levels that appear to have occurred.

The PCB lab reports were reviewed to determine if lab variance or errors could account for these cycles. There was no obvious lab errors found that could account for the observed changes. However, as noted in the MW2 I test report, it is possible that the length of time that any particular sample sat in the autosampler could have influenced the result. As in that test, samples that sat in the sampler for longer periods of time tend to have lower levels of PCBs. This could explain the apparent cycling levels of PCBs.

One of the main goals of the pump test was to determine if significant levels of PCBs would be drawn from this well while pumping and if the PCBs at the IC spring would fall during the pumping period. While there are some unexplained drops of PCB levels at IC Spring during the pumping period, these drops are not consistent or large enough to consider them a direct benefit of pumping at this well. Additionally the level of PCBs withdrawn from the well dropped as pumping began and stabilized at a low level that would not indicate a significant impact on PCBs at the spring would be expected.

3.2.4 Dye Response at the Pumping Well

Dye was sampled at the pumping well to determine if any residual from the dyes previously injected in the aquifer on October 31 would be drawn to the well while pumping. If and when this occurred, may provide an indication of how far the well is from a flow path carrying the dyes. The dye sample results are shown on Table 12.

As can be seen on the table, at the very beginning of pumping there were no significant levels of any of the recently injected dyes. After 8 hours of pumping Fluorescein begins to show up at low levels. These levels build a breakthrough curve with a peak 28 hours after pumping began. The magnitude of the Fluorescein levels are low relative to those recovered at MW'I or IC Spring. Speculation is that this Fluorescein is coming from a continual slow release of dye from the LF6 conduit. This slow release hits the part of the phreatic zone influenced by this pumping well at some location, the travel time to the well of 8-28 hours may be indicative of the distance to that location. Figure 15 shows the Fluorescein results at this well.

Sporadic hits of cosine are also periodically seen after 8 hours of pumping. Curiously, there is not a breakthrough curve for cosine after the NN700 well is flushed on November 15 at 11:30. However, the cosine is more consistently present for the next 20 hours after the flush. No phloxine is detected in the well even after SPI was flushed on November 14.

A second low level breakthrough curve for Fluorescein is seen starting 8 hours after LF6 is flushed on November 16. This indicates that at least some of the Fluorescein flushed is captured by this well.

These results indicate the following:

• Phloxine does not take a path from SP I to the IC Spring which is influenced by this well.
• The cosine pathway is somewhat influenced by this well but not to a significant extent.
• Minor portions of the Fluorescein pathway can be influenced by this well.

Overall, these results indicate this location is less efficient than MW2 I for capture of site area waters and PCBs.

3.2.5 Dye Response at the IC Spring

Dye samples at IC Spring were taken throughout the MW16 pumping period. See Table 11 for a complete listing of the results. Only Fluorescein was reliably and consistently detected at the spring. These results are also shown on Figure 16.

The analysis of the Fluorescein response to pumping at the spring is confounded by the fact that the Fluorescein concentrations at the spring were receding from the October 31 dye injection and again from the November 8 flushing of LF6.

As can be seen from Figure 16, Fluorescein concentrations during pumping are slightly higher for the first day then they were immediately prior to the start of pumping. They then continue a typical receding limb decline with a couple of perturbations until the large breakthrough response on November 19. The breakthrough curve starting at 12:00 on November 19 is attributed to the flushing at LF6 that occurred on November 16 at 15:30. This is a travel time of about 68 hours. With flows at IC Spring between 30 and 40 8pm during this time period, an expected travel time of 50 to 65 hours was expected. This is fairly good agreement considering LF6 is an epikarst location.

Comparing the size of the breakthrough curve at the spring to the breakthrough at MW16 from the November 16 flush shows that by far most of the Fluorescein flushed at LF6 was not captured at the pumping well. Also note that the Fluorescein response at IC Spring on November 19 is a two peak response. This compares with the original Fluorescein response from the October 31 dye injection. Recall that the Fluorescein response from the November 8 flush (while pumping MW21) was a single peak at this spring.

Table 13 shows a summary comparison of dye results obtained for each of the Fluorescein recoveries resulting from the October 31, November 7 and November 16 flushes.

3.2.6 Conclusions and Recommendations for MW16 Test

It appears that pumping MW16 cannot significantly lower the PCBs at IC Spring. It also appears that pumping this well will not capture most of the materials that are flushed from the epikarst near LF6.

Hydraulic indicators such as low flow water levels, quick equilibration and dye recovery during pumping (along with relatively low drawdown per gallon/minute pumped) show that this well is well connected to a major conduit but not as close to the conduit as MW21. The conduit affected most likely drains the 795-800 elevation zone from the southern portion of the site.

3.3 Analysis of Drawdown Data for MW21 and MW16 Pump Tests

Table 5 shows the data taken for MW21 while it was the pumping well, and Table 9 the data for MW16 while it was the pumping well. Table 14 shows the data for the spot observations measured by hand.

There are two ways of looking at the drawdown data from the pump tests. One is to perform time-drawdown analysis and the other is to perform distance-drawdown analysis.

Time-drawdown analysis is based on the Jacob Assumption for the Theis Non-equilibrium theory. This is the analysis of the observation wells during pumping as the levels are declining before they reach maximum drawdown or equilibrium. The Theis theory assumes:

• The transmissivity of the aquifer tapped by the pumping well is constant during the test to the limits of the cone of depression.
• The water withdrawn from the aquifer is derived entirely from storage and is discharged instantaneously with the decline in head.
• The discharge well penetrates the entire thickness of the aquifer, and well bore storage is negligible.

The Jacob Assumption is that the data in a semi-log graph of drawdown vs. time (with drawdown on the linear axis and time on the logarithmic axis) will plot as a straight line, once stead-shape conditions have occurred in the cone of depression. This means that the slope of the line is proportional to the pumping rate and the transmissivity of the aquifer. Departures of the data plot from a straight line may give information about changes in aquifer properties or proximity to boundary conditions.

Distance-drawdown analysis makes the same assumptions as the Theis theory. Comparisons can be made during pumping if measurements are made simultaneously. If withdrawals have occurred long enough that increases in recharge andJor decreases in discharge have balanced to the rate of withdrawal and drawdown has ceased, then cone of depression in said to be in a steady state. Resumption of drawdown or recovery of levels, if pumping rate remains constant, reflect the intersection of the cone with boundary conditions. Again, the Jacob Assumption is that the data in a semi-log graph of drawdown vs. distance from the pumping well (with drawdown on the linear axis and distance on the logarithmic axis) will plot as a straight line. The slope of the line is proportional to the pumping rate and the transmissivity of the aquifer. Departures of the data plot from a straight line may give information about changes in aquifer properties or proximity to boundary conditions.

Figure 17 is a semi-log of drawdown vs. time for the MW21 pump test and Figure 18 is the plot for the MW 16 test. In the MW21 test, drawdown levels off at 23-25 hours, and then begins again at a higher rate. If this was not an increase in pumping rate (and field notes do not indicate any increase) then the cone of depression reached a less transmissive part of the aquifer. Because this happens nearly simultaneously in all observation wells, we might speculate that the upstream end of the branchwork system of conduits had been reached. Drawdown then appears to reach steady state at 36 hours. The recovery at 49 hours probably indicates a decrease in pumping rate.

In the MW16 test the pumping rate was slightly less (16.8 vs. 17.0 8pm) and the wells appear to begin to reach equilibrium at 21 hours. At 27.5 hours, 47.5 hours, and 76 hours water was flushed into wells NN700, SPI, and LF6-8 respectively, during the MW16 test. The water flushes into the phreatic wells (NN700 and SP I ) seems to have caused an instantaneous rise in water levels in all wells. The water flushes into LF6-8 during the MW16 and MW21 tests does not seem to have had much of a discernable effect. A very small rise in NN300A on November 16 at 16:30 may have been due to the LF6-8 flush, but it is too small to be definite. The effect of a mere 100 and 300-gallon flush indicates the aquifer has very small storage capacity. In the MW 16 test, as in the MW21 test, drawdown continues after the 21- hour hiatus, perhaps indicating a discharge boundary has been passed. Drawdown then appears to reach steady state at about 83 hours. There is a slight recovery at 94 hours, again which may be due to a decrease in pumping, and then steady state is achieved again at 105- 106 hours, albeit at a slightly higher level.

The slopes of the lines on Figure 17 and 18 are all remarkably parallel. This indicates that the zone being pumped has a relatively homogeneous transmissivity. However, there appears to be a greater degree of drawdown from MW 16, despite the nearly equal pumping rates. This can be seen in Figure 19, which compares drawdown from observation wells similar distances from their respective pumping wells, and Figure 20, that compares drawdown in MW8s, nearly equidistant from both pumping wells. Figure 21 shows the drawdown in the pumping wells. Until about 240 minutes (4 hours) MW16 is drawing from a less transmissive zone than MW21. Our interpretation is that MW16 is drawing water more from the anastomotic zone at the 795' to 800' elevation, while MW21 is drawing water from a nearby larger conduit. That is why pumping from MW16 draws the anastomotic zone down almost 50% more than MW21.

Figure 22 is the plot of maximum drawdown vs. distance for the MW21 test and Figure 23 is the plot of maximum drawdown vs. distance for the MW16 test. For the MW21 test, SP I and MW8s appear to show less drawdown than would be expected. For the MW 16 test MW 17 does not seem to show the same relative drawdown as well as SP 1 and MW8s. Figure 24 is a plot of all phreatic wells measured on November 16 (between 70 and 73 hours) during the MW16 test. That plot shows particularly large deflections from the straight line plot of data for the 4-series wells (20, 4I, 21, 4s). This phenomenon is manifested on Figure 18 in the order of relative drawdown. The closed well to MW16 (with datalogger) was MW 15, which has the largest degree of drawdown, the next closed well is MW17, but MW6, NN625, and MS 1 have greater degrees of drawdown.

So the hydraulic data is showing that the anastomotic zone in the 795-800' elevation is widespread and relatively uniformly dissolutioned. However, there is another network of conduits, which, while still in connection with this anastomotic zone, are not well connected with each other. On Figure 22 MW7 appears to show expected drawdown levels when MW21 is pumping. But on Figure 23, MW17 appears as an anomaly when MW16 is the pumping well. This indicates that MW17 is on, or closer to, a different network of conduits than MW16. Based on Figure 24, our interpretation is that the 4-series wells are near a major branch conduit. It should be noted that the 4-series wells are the potentiometric low point in base flow periods, and that the low flow order is the same as the deflection in Figure 24 i.e. MW20 > MW4I > MW21 > MW4s.

4.0 Sargent's Pond Injection Test

The Sargent's Pond injection test consisted of releasing the water stored in the lined retention pond to Sargent's Pond. This water came from the previous pump test (after treatment for removal of PCBs). There was approximately 200,000 gallons of water stored from the pump tests. The goal of the test was to see how the water once released would impact the pond level, well levels around the site and flow/PCBs at IC Spring.

Prior to release, continuous well level monitors were installed at both phreatic and shallow wells/piezometers near the site. Also an autosarnpler was set to take PCB samples at IC Spring. The following locations were continuously monitored:

NN300A, SP1, PZD, PZBD, B3, MW6, MW15, MW17, NN625, MS1, Sargents Pond Level and IC Spring flow/conductivity

The release occurred on November 20, 2001 at 13:30. The phreatic well levels are shown on Tables 15 and 16. The shallower well and Sargent's Pond data are shown on Table 17. IC Spring data is shown in Table 18.

Refer to Figure 25 for a view of the phreatic well data. Figure 26 shows the shallower well and Sargent's Pond data and Figure 27 the spring data.

The most obvious result of this testing is that there was not a large impact on well levels or spring flow. This is confimned by the rise and then negligible fall in Sargent's Pond level after the release. This shows that the pond has a low leak rate in this level range.

The flow records at the IC Spring show what a small rise in flow occurs after release of the water to the pond. The rise is approximately 4-5 8pm.

The PCBs at IC Spring were sampled every two hours until November 23 10:00. Refer to Figure 28. There is a large apparent drop in PCBs beginning on November 21 at 12:00. While this could be do to dilution water added to the aquifer from the injection, it is more likely an artifact of sampling. Again, the samples taken on November 21 from 12:00 on spent more than 24 hours in the autosampler and are most likely biased low. The results of November 23 which spent little time in the sampler are most likely liKle influenced by sampling artifacts. These show a lowering of PCBs also from pre-injection levels.

5.0 Overall Conclusions (Note this includes all tests)

ù Material flushed from LF6 appears to take two pathways from the site to the spring. ù It appears that pumping MW21 was much more successful than MW16 at intercepting the lesser of these two pathways.
• One major pathway from LF6 is into Valhalla Cemetery while still in the epikarst. Some of the epikarst waters descend to the phreatic zone near well 00370.
• Wells on the east side of the landfill at the tested depth are not effective at intercepting flushed PCBs from the epikarst.
• The dye results suggest that capturing PCBs as they are flushed may be most effectively done either in Valhalla Cemetery, or at the extreme southern portion of the landfill up against the railroad tracks.
• These results also suggest that dewatering the epikarst may be effective at lowering both non-storm flow and storm pulse PCBs to the spring.
• Most of the materials leaving the southeast corner of the site in the epikarst must be descending to a portion of the phreatic zone that was not impacted by either pump test. Additional investigations are necessary if these zones are to be found.
• Some of the PCB results at IC Spring are questionable because the samples sat in the sampler for longer periods of time. This is most likely caused by volatilization of PCBs from the uncapped bottles as the sampler warmed up in the sun. The weather was unusually warm during November testing. Future testing should take precautions against this sort of potential error such as icing the sampler and minimizing the time that a sample remains in the sampler to less than 24 hours.
• Sargent's Pond has a low leakage rate for elevations up to 843 feet arnsl.

Footnotes