Products Services Classes Downloads  FAQ  Contact Home


The Effectiveness Of Intrawell Ground Water Monitoring Statistics At Older Subtitle D Facilities   

Henry R. Horsey, Ph.D.
Phyllis Carosone-Link, M.S.
Megan R. Sullivan
Jim Loftis, Ph.D.


The reality of the new statistical evaluation requirements for ground water at Subtitle D facilities are being felt nationwide. All too often these requirements are resulting in unnecessary increases in ground water monitoring costs and erroneous indications of facility impacts. For example, a common situation faced by many older facilities is that there is statistically significant natural spatial variation in the hydrogeology underlying their site, but their ground water monitoring began after waste had been placed. If the standard approach of comparing water quality between upgradient and downgradient wells is used (interwell analysis), the statistical tests will often be unable to distinguish between a statistical difference in water quality caused by a release from the facility and a statistical difference caused simply because of the natural upgradient versus downgradient hydrogeologic variation. Consequently, where there is statistically significant variation in the natural hydrogeology an intrawell analysis is the preferred statistical approach. 

The intrawell analysis is fundamentally different from the interwell analysis. While the interwell analysis compares compliance wells against a background composed of upgradient well data, the intrawell analysis compares each compliance well against a background composed of it's own historical data. When an intrawell analysis, however, is used in a detection monitoring program, the implicit assumption is that the historical compliance well data that is used as background has not been impacted by the facility. The problem faced by older facilities is that the ground water monitoring began after waste had been placed at the facility. How do they demonstrate that their historical compliance well background data are "clean" and thus an intrawell analysis is appropriate? 

Several alternative statistical analysis and hydrogeologic investigation approaches have been developed to demonstrate that historical compliance well background data have not been impacted by a facility. An analysis of the effectiveness of these approaches at sites where they have been implemented over the last two years is very promising. These approaches have resulted in intrawell limit based tests that provide an earlier indication of a change in ground water quality than statistical trend tests without the high false positive rates of interwell analyses. There are situations, however, where the intrawell approaches have resulted in limits that are so stringent that facilities may find it more advantageous to remain with the hydrogeologically incorrect interwell statistical tests. This paper will discuss the effectiveness of the intrawell statistical approaches at four Subtitle D facilities where waste had been placed prior to the inception of the ground water monitoring program. 


At older solid waste facilities, there is often significant natural spatial variation and waste that has been placed prior to the inception of the ground water monitoring program. This situation creates a particular dilema for these facilities when they begin to implement ground water quality statistical analyses. Should they choose to use interwell analyses, they substantially increase the chance of being unnecessarily forced into assessment monitoring or corrective action. Yet, before they can use intrawell analyses, they must convince skeptical regulators that the historical data in their compliance wells do not show facility impacts. 

The difficulty with interwell analyses is that they require that the hydrogeologic conditions are contiguous and uniform throughout the site and that all wells are screened in the same interval. These conditions must be met or natural differences in constituent levels can confound the background-to-compliance comparisons and may lead to erroneous conclusions about facility impacts. Thus, a significant disadvantage of interwell tests is their decreased power to detect differences in the presence of spatially varying hydrogeology. 

Another disadvantage of some types of interwell tests is that they require the pooling of compliance data over time to meet minimum compliance sample requirements and thus are subject to a dilution effect when data are pooled across reporting periods. It is this concern with analysis of variance (ANOVA) tests that has lead the California State Water Resources Control Board to require large sample sizes within a single reporting period (i.e., four to nine per well) when ANOVA is used. Unfortunately, collecting multiple samples within a single reporting period compromises the statistical requirement that the samples be physically independent. A lack of sample independence leads to reduced variability and ultimately to increased false positives. 

The intrawell analysis does not require that the hydrogeologic conditions are contiguous and uniform throughout the site and that all wells are screened in the same interval. This is because the intrawell analysis identifies changes over time at a given well instead of changes between wells. The difficulty with intrawell analyses when used in a detection monitoring program, however, is that the historical data used to develop the background limits can not have been impacted by the facility. If this condition is not met, then the limits will not necessarily signal when a facility impact occurs in the future. Thus the power of the test is compromised. 

The ideal situation in which to implement an intrawell analysis is at a new facility prior to the placement of any waste. Then a number of samples can be taken at each of the compliance wells prior to waste placement. These samples can be used to develop the intrawell limits and there can never be a question as to whether the samples have been impacted by the facility since no waste had yet been placed. Implementing intrawell analyses at older solid waste facilities is a far less than ideal situation. Usually there are no pre-waste data available at the compliance wells. The challenge in using intrawell analyses at these sites then becomes one of demonstrating that the prospective intrawell "background" data haven't been previously impacted by the facility. This task becomes particularly more onerous when downgradient levels are significantly higher than upgradient levels. 

One methodology to justify the use of intrawell analyses at older sites around the country consists of utilizing intrawell historical data that have been screened for high statistical outliers, temporal trends, seasonal effects, and the presence of anthropogenic compounds. The screening process eliminates any observations that could possibly be suspected of representing a prior facility impact from the pool of prospective background data. 

This approach has been used at numerous sites around the country. As additional data becomes available, the appropriateness of this approach has been reinforced. The following section describes four sites where this approach to develop intrawell limits was used. The effectiveness of these limits is then discussed using recent monitoring data. 

Case Study Sites


sample screen of interwell statistical analyses

Site A. This site provides a textbook example of a landfill where the use of interwell statistical analyses are inappropriate because of heterogeneous hydrogeology underlying the site. At this site there is an intrusion of a saline lake into the freshwater aquifer underlying the site downgradient of the landfill. Downgradient well concentration levels of general chemicals that are common to saline water (boron, calcium, chloride, magnesium, manganese, molybdenum, potassium, sodium, sulfate, and TDS) were significantly higher than the upgradient concentration levels. Interwell analyses for detection monitoring at this site would have needlessly resulted in proceeding to an assessment monitoring program when, in fact, upgradient-downgradient differences were due to naturally occurring spatial variation. In fact, eleven out of the thirteen monitoring parameters had statistically significant findings when interwell analyses were used (see figures 1 and 2).

sample chart of parametric interwell tolerance limit 


sample chart of non parametric inter well prediction interval 

Because volatile organic compounds (VOCs) were all below detection levels and inorganic constituents of concern fell below background limits, we were able to demonstrate that a facility impact was not the cause of the elevated inorganic chemistry parameters. 

To ensure regulatory acceptance of the intrawell background data for this site, we also screened these data for high statistical outliers, temporal trends and seasonal effects in addition to the VOC analysis. The resulting intrawell prediction limits were used as background standards for one year. At the end of that year the background data were reassessed to determine whether the compliance observations could be consolidated with background to derive the next year's limits. 

sample screen of contrast test indicates 

Site B. A significant problem with naturally occurring spatial variation is that it is present at many sites, but is not always detected. For example, there was very limited hydrogeologic information available when site B began collecting samples for detection monitoring. Since sampling had just begun, the background sample size was small. For this reason an interwell analysis method, that has a small minimum background sample size requirement, was chosen for detection monitoring. As the background sample size increased over time and the background limits became more precise, a statistical exceedance was found for chloride. Instead of precipitously moving into assessment monitoring, a statistical investigation was initiated to determine if naturally occurring spatial variation, as opposed to a facility impact, was responsible for the statistical differences seen between upgradient and downgradient concentration levels. 

The multiple background wells were compared and statistically significant differences were found among them. In fact, one background well had higher concentration levels than the downgradient well. The multiple well comparison as well as the absence of anthropogenic compounds in the downgradient well strongly suggested that the statistical difference seen was due to naturally occurring spatial variation at this site as opposed to a facility release. Consequently, interwell-types of analyses were inappropriate for detection monitoring at this site. 

By the time this determination had been made there were sufficient data for an intrawell type of analysis. There were no trends or outliers in the downgradient data. Thus, all of the available data were used to establish intrawell background limits. 
sample graphic of discontinuous aquifer 

Site C. This site overlay fractured bedrock, resulting in a discontinuous aquifer; thus an interwell analysis was inappropriate. The exploratory data analysis at this site showed an absence of anthropogenic constituents, temporal trends, and seasonal effects. 

sample chart of non parametric inter well tolerance limit 

Although upgradient concentration levels were similar to downgradient levels for most wells and constituents, there were a few exceptions in which interwell analyses resulted in statistically significant findings. Had compliance at this site been determined by interwell analyses, this site would have needlessly incurred the added costs of retesting and/or assessment monitoring. 
sample graphic of a landfill site geology over upgradient and downgradient well 

Site D. This site's geology consisted of a laterally discontinuous silt and sand bed. Consequently, an interwell analysis was inappropriate. Upgradient concentration levels were similar to downgradient levels for most wells and constituents, however there were a few exceptions in which an interwell analyses resulted in statistically significant findings. 

sample screen capture of contrast test 

The exploratory data analysis at this site showed an absence of anthropogenic constituents, temporal trends, and seasonal effects. This would suggest that an intrawell approach was appropriate. Nonetheless, a more regulatory conservative screening approach was required for regulatory acceptance of intrawell limits at this site. Consequently, a second phase screening was implemented and a tolerance limit generated from upgradient well data was used to screen prospective background data for the downgradient wells.

The use of an upgradient based tolerance limit to screen prospective downgradient data is questionable from a statistical perspective. This approach is essentially an interwell test in that it presupposes that upgradient limits are relevant for screening downgradient background data. We know that the site's significant spatial variation suggests that an interwell approach is not appropriate. Thus, there is a legitimate concern that the long-term consequence of using this interwell screening method will be to increase the false positives and thus monitoring costs at this site. Nonetheless, this approach did not result in any initial statistically significant findings while the normal interwell approach did. 


Effectiveness Of The Intrawell Analyses
Since intrawell statistical analysis tests were approved for the detection monitoring programs at the previously described sites, a number of quarterly samples have been collected. We retrospectively analyzed this data to investigate the effectiveness of the intrawell approach in detecting facility impacts to the ground water and in minimizing false positive findings. 

Site A. Quarterly monitoring at this site has been ongoing with little change to ground water quality since intrawell background limits were established. None of the intrawell limits have been exceeded and, in fact, concentration levels appear to be decreasing over time (see Figure 4) for one of the wells. By contrast, when interwell tests were run on this data numerous statistically significant findings occurred because of the interwell's inability to discern between differences caused by spatial variation and differences caused by a facility impact. 

Because the natural concentration levels are changing at this well for at least one constituent, that constituent's intrawell limit at that well will be updated on a biannual basis to ensure that the limit will retain the power to detect a future facility release. 

Site B. All constituent concentration levels have remained below their respective intrawell limits. Again, interwell tests resulted in numerous false positive findings. Trend tests indicate no change in water quality. 

Site C. No intrawell statistical exceedances have occurred at this site to date. Furthermore, concentration levels appear to be decreasing at one of the wells for some constituents as evidenced by a significant decreasing temporal trend (see Figure 5). As a result of the trend analyses, the relevant background limits at that well will be revised (i.e., lowered) for these constituents so that they reflect more current conditions. 

Site D. Two consecutive quarterly samples have exceeded intrawell limits for chloride. Although three out of four verification retests have not confirmed the statistically significant findings, trend tests indicate an increasing temporal trend for chloride. 

A Mann-Kendall trend test that was performed on the chloride data collected as of October 1995 was significant at the 0.05 confidence level for the most recent sampling period (see Figure 6). The intrawell prediction limit analysis detected a statistically significant finding two quarters prior to the significant trend test. Thus, this indicates that the background data two-phase screening methodology used at this site results in very sensitive background limits for this facility. 

We then analyzed the effectiveness of intrawell limits based upon background data that had not been screened using a tolerance limit in identifying changes in water quality concentrations. The resulting limit for chloride was as effective as the two-phase screening limit in identifying the statistical exceedance. This would suggest that the second phase screening did not really result in better protection of the ground water. 


Statistical analyses is one tool that geologists and engineers use to determine water quality at a facility. Statistical tests are not intended to be used in isolation of other types of meaningful evidence about a site. It is important to use statistical analyses in the context of the site's hydrogeology to obtain meaningful results. 

The use of interwell comparisons for detection monitoring is not appropriate when hydrogeologic variation is naturally present at a site. An alternative type of analysis, intrawell analysis, may be difficult to justify at sites where waste has been placed prior to water quality monitoring. 

We have developed a methodology to justify the use of intrawell analyses at a number of sites around the country where waste was in place prior to ground water monitoring. The basic screening method consists of: 

Screening for the presence of VOCs; and 

Removing high statistical outliers from intrawell background data 

The effectiveness of using this methodology has been described in this paper. Evidence of the effectiveness includes lower false alarms that would have resulted from interwell analyses, incorporation of well-specific hydrogeology, and the needed sensitivity to detect a facility impact. Thus, this approach: 

Minimizes site-wide false positives; 

Is appropriate for the site hydrogeologic conditions; and 

Maximizes statistical power to detect a facility impact. 

When permitting agencies require extraordinary measures to ensure very conservative intrawell limits, additional screening steps may be used to achieve regulatory acceptance of the intrawell limits. However, we have serious concerns about the consequences of using these more stringent screening methods which violate hydrogeologic assumptions. 

  © Copyright 2006 Sanitas Technologies. All Rights Reserved.