Consider updating your groundwater monitoring program to reduce the occurrence of statistical false positive and negative rates
RCRA groundwater monitoring regulations have a few key goals: Accurately characterizing existing groundwater quality at the facility in question, assessing whether a hazardous constituent release has occurred and, if it has, determining whether measured levels meet the compliance standards. If you use accepted statistical testing methods in your groundwater monitoring program, you should be able to make the correct decisions about a facility’s regulatory status.
Now let’s dig deeper into what that means and how you can get there.
Compliance with what?
Groundwater monitoring requirements are described in Title 40 CFR Part 258, Subpart E – Ground-Water Monitoring and Corrective Action. States also typically have additional regulations. There is always the need to balance false positives and false negatives in statistical testing procedures. To help users navigate this issue, EPA issued “Unified Guidance (USEPA, Statistical Analysis of Groundwater Monitoring Data at RCRA Facilities, March 2009). There’s a lot at stake here – the guidance is meant to minimize the risk of falsely declaring a site to be out-of-compliance or of missing real evidence of an adverse change in the groundwater.
Three main steps in a groundwater monitoring program
Detection monitoring is required at nearly all municipal solid waste landfills (MSWLFs), using a groundwater monitoring system made up of monitoring wells placed upgradient and downgradient of the MSWLF. At a minimum, the program must include monitoring for the 62 constituents listed in Appendix I of the 40 CFR Part 258.
Assessment monitoring is required whenever a statistically significant increase (SSI) over background has been detected for one or more of the constituents listed in Appendix I of the regulations. Within 90 days, Appendix II parameters (214) need to be analyzed. Minimum of four independent samples from each background/downgradient well must be collected and analyzed to establish background for the new constituents.
Corrective action must be taken within 90 days of finding that any of the constituents listed in Appendix II have been detected at a statistically significant level exceeding the groundwater protection standard (GWPS).
It is important to consider the difference between comparison of the compliance data point to the background data or to the GWPS. An analyte can be identified as an SSI but may be well below the GWPS! If there is an SSI, you should take a closer look at this specific well-constituent pair. Also, consider doing a review of the parameter’s results at other wells and/or review other parameters at the well (for example, leachate indicators). Ask yourself, “Where is the well located? Is there a trend observed at the upgradient wells?”
What can you do to strengthen groundwater monitoring statistics?
Groundwater detection monitoring can involve either inter-well analysis, which compares groundwater quality at the downgradient wells to the upgradient wells or intra-well analysis, which compares groundwater quality at the same location over time.
Inter-well statistical tests are best conducted when there is a lack of spatial mean differences and a common variance; and if we can assume that changes in groundwater quality only affect downgradient (compliance) wells and not upgradient (background) wells.
Intra-well tests are suitable when there is evidence of groundwater mounding or other anomalies that could lead to the lack of a reliable, definable gradient; if there are changes that affect background wells, like seasonal fluctuations; or if there is evidence of spatial variation (natural or man-made differences in mean levels) observed among wells known to be uncontaminated (i.e., upgradient background locations).
Statistical tests used to evaluate groundwater monitoring results require appropriate and representative background measurements. EPA’s Unified Guidance recommends obtaining a high number of background observations. Updating the groundwater monitoring program to increase the participant size helps decrease the occurrence of statistical false positive and negative rates.
The Unified Guidance recommends collecting a minimum of 8 to 10 independent background observations before running most statistical tests and updating background information when at least 4 to 8 new measurements are gathered (every 1-2 years in case of quarterly sampling) and with semi-annual sampling, every 2-3 years.
Does your background include outliers? If you do not remove an outlier value with much higher concentration than other background observations prior to statistical testing, you will tend to increase both the background sample mean and standard deviation, which may also substantially raise the magnitude of a control limit calculated from that sample. If this happens, a subsequent compliance result will be much less likely to identify a SSI. Or, if the maximum is an outlier not representative of the background population, few truly contaminated compliance wells are likely to be identified by such a test, thereby lowering the statistical power of the method and the overall quality of the statistical monitoring program.
Another issue is what to do with trending data – should it be added to the existing background data set? An increasing or decreasing trend may be apparent between the existing background and the newer set of candidate background values, either using a time series plot or applying a trend test. Most detection monitoring tests assume that background is stationary over time, with no apparent trends or seasonal variation. A mild trend can make a very little difference, while more severe or continuing trends are likely to be flagged as SSIs.
With inter-well tests, a stronger trend in the upgradient background well(s) may signify a change in natural groundwater quality across the aquifer or an incomplete characterization of the full range of background variation. If a change is evident, it may be necessary to delete some of the earlier background values from the updated background sample, to ensure that compliance testing is based on current groundwater conditions and not on outdated measures of groundwater quality.
Selecting the right statistical methods
The statistical methods used to evaluate groundwater monitoring data have to be protective of human health and the environment – and appropriate for the data distribution of parameters. If the normal distribution of the parameters cannot be achieved, then the data may need to be transformed. If the distributions for the constituents differ, more than one statistical method may be needed.
Prior to selecting the test, you must determine the statistical parameters after considering the number of samples in the background water quality database, the sampling frequency, and the range of the concentration values for each constituent of concern. Here are a few examples of the most useful tests out there for groundwater monitoring:
The box and whiskers plot is a useful detection monitoring method for visualizing the variation within and among wells. When variation is present among upgradient wells, intra-well analyses are generally recommended. If the upper box and whisker are approximately the same length as the lower box and whisker, with the mean and median approximately equal, then the data are distributed symmetrically. If one box (upper or lower) and whisker are longer than the opposite box and whisker, then the data are skewed.
Also useful are tolerance limits, an inter- or intra-well analysis that compares compliance observations to a limit established by background data that is constructed to contain a specified proportion of the observations population (e.g., coverage of 95 percent).
The prediction limit test is an inter- or intra-well analysis that compares one or more compliance observations to a limit established by background data. These limit-based tests are recommended for sites in detection monitoring to determine whether changes are occurring at compliance wells.
Shewhart-CUSUM control charts measure both rapid releases and long-term, gradual trends within a given well for a given constituent. These tests use screened background data from within a well to establish a baseline for comparison of future data.
Finally, the Mann Whitney (Wilcoxon-Rank Sum), an inter- or intra-well test is a nonparametric test that may be used to evaluate whether the median measurement from one population is significantly higher or lower than the median measurement from another population. This test is useful for updating the background data set (i.e., comparing the median of the current background data set to the median of the candidate background).
Update your program to ensure success
To sum it up, updating the groundwater monitoring program to increase the participant size will help decrease the occurrence of statistical false positive and negative rates. In addition, exclusion of potential outliers and other trending values, as well as a higher number of background observations, contributes to a successful background update.
When was the last time background was updated at your facility? Can your facility revert back to detection monitoring?
Magda Mendola is a project manager holding degrees in Sanitary and Environmental Engineering. She has more than thirteen years of experience primarily with groundwater and soil remedial design, data management and compliance.
Categories: Environmental Planning & Compliance, Environmental Remediation, Landfill Engineering and Design, Solid Waste
Posted By Magdalena Mendola at 11:30 AM | No Comments on Consider updating your groundwater monitoring program to reduce the occurrence of statistical false positive and negative rates
Leave a Reply
Your email address will not be published. Required fields are marked *