Groundwater Statistics and Interpretation at Landfills It can be a - PowerPoint PPT Presentation

Groundwater Statistics and Interpretation at Landfills It can be a useful tool . . . honest! Paula Leier-Engelhardt, P.G., C.P.G. Principal Geologist HydroGeo Solutions LLC

• How do you get the data in the first place • Create and maintain a database • What is the purpose for the statistics, and what methods should be used 4/26/2017 2 U.P. Solid Waste Forum, Marquette MI

The Groundwater Data • Your statistics will only be as good as the data you are using • Sampling team • Proper training and equipment • Laboratory • QA/QC • Review detection limits; ask questions if they change 4/26/2017 3 U.P. Solid Waste Forum, Marquette MI

The Groundwater Data (cont.) • Upon receiving analyses, review and compare to historical results • Can catch the obvious things • Detection limits • Were all the parameters analyzed? • Is there a sampling point missing? • Pay special attention to the field data 4/26/2017 4 U.P. Solid Waste Forum, Marquette MI

Create and Maintain a Database • Should be in a program separate from a statistical software package, such as Excel or Access • What goes in it? Everything! • Do not change anything from the original lab submittal (except to correct obvious errors) 4/26/2017 5 U.P. Solid Waste Forum, Marquette MI

The PQL vs. MDL vs. Reporting Limit Dilemma CAS Date Site Client Collection Reporting Qualifier Analytical Analysis Prep Date Registry Substance Name Result Units Analyzed Identifier Sample ID Date Limit Code Method Comments Number XXXXX MW-2 2/14/2017 ALK-CACO3 alkalinity (as caco3) 330000 ug/L 2000 2320 B-1997 2/17/2017 2/17/2017 973577 2/15/2017 2/15/2017 XXXXX MW-21 2/14/2017 SW301 chlorides 20000 ug/L 2500 SW846 9056 973569 2/15/2017 2/15/2017 XXXXX MW-4A 2/14/2017 SW301 chlorides 2600 ug/L 2500 J SW846 9056 973574 Reporting limit: Constituent concentration that, when processed through the complete method, produces a signal that is statistically different from a blank. 4/26/2017 U.P. Solid Waste Forum, Marquette MI 6

What is the purpose of the statistical analyses? • R 299.4908 -- Landfill groundwater monitoring; statistical procedures • Release detection • Are concentrations greater than background concentrations, or above/below a criterion? • Is the sampling frequency appropriate? • Is the monitoring network appropriate? U.P. Solid Waste Forum, Marquette MI 4/26/2017 7

Statistical Methods for Release Detection • Tolerance Intervals • Prediction Intervals • Shewhart/CUSUM Control Charts U.P. Solid Waste Forum, Marquette MI 4/26/2017 8

Tolerance Intervals • Statistical ranges constructed from on-site background data. • Tolerance limits define the range of data that fall within a specified percentage with a specified level of confidence. • An upper tolerance limit (UTL) is designed to contain, but not exceed, a large fraction (that is, 95%, 99%) of the possible background concentrations, thus providing a reasonable upper limit on what is likely to be observed in background. U.P. Solid Waste Forum, Marquette MI 4/26/2017 9

Tolerance Intervals • Assumptions Data are stable; no trends; no seasonal variation Little spatial variation between background and compliance monitoring wells Background data are normally distributed, or transformed normally distributed U.P. Solid Waste Forum, Marquette MI 4/26/2017 10

Tolerance Intervals Strength Weakness • Built-in ‘failure rate’ • Fairly easy to calculate • The more non-detects you have, • Best used where you can pool the more data you need all the data from the upgradient • It can be used for intrawell wells to create your tolerance limits analyses, but the statistics start to fall apart at a sample size greater than 50 U.P. Solid Waste Forum, Marquette MI 4/26/2017 11

Statistical Methods for Release Detection • Tolerance Intervals • Prediction Intervals • Shewhart/CUSUM Control Charts U.P. Solid Waste Forum, Marquette MI 4/26/2017 12

Prediction Intervals • A prediction limit (or interval) is used to determine whether a single observation is statistically representative of a group of observations. • The interval is constructed from a background set of observations such that it will contain K future compliance observations with stated confidence. • If any observation exceeds the bounds of the prediction limit, this is statistically significant evidence that that observation is not representative of the background group. U.P. Solid Waste Forum, Marquette MI 4/26/2017 13

Prediction Intervals • Assumptions Background data are normally distributed, or transformed normally distributed Data sets with seasonal variations, or spatial variations, can be used, but adjustments must be made At historically contaminated compliance wells, establishing a proper baseline for a prediction limit is problematic, since uncontaminated concentration data cannot be collected. U.P. Solid Waste Forum, Marquette MI 4/26/2017 14

Prediction Intervals Weakness Strength • Quite flexible – EPA describes • Re-testing must be done nine different prediction limit test • Nonparametric prediction limits options typically require a much larger • A prediction limit estimates a sample size than parametric firm ‘cap’ on the background prediction limits to achieve the population for a specified desired confidence level. number of future sampling events (comparisons). Allows for clear interpretation of when background levels have been exceeded. U.P. Solid Waste Forum, Marquette MI 4/26/2017 15

Statistical Methods for Release Detection • Tolerance Intervals • Prediction Intervals • Shewhart/CUSUM Control Charts U.P. Solid Waste Forum, Marquette MI 4/26/2017 16

Shewhart/CUSUM Control Charts • Commonly used to monitor the stability of groundwater data and to detect changes in data trends that may require further investigation. • The Shewhart control limit tests for and flags a sudden spike or change in trend of the data, which may indicate an event such as a new release at the site. The CUSUM control limit tests for and flags a gradual, but significant, increase or decrease over time, which may, for example, indicate plume migration. U.P. Solid Waste Forum, Marquette MI 4/26/2017 17

Shewhart/CUSUM Control Charts • Assumptions Background data are normally distributed, or transformed normally distributed No trends are present in the background data U.P. Solid Waste Forum, Marquette MI 4/26/2017 18

4/26/2017 19 U.P. Solid Waste Forum, Marquette MI

Shewhart/CUSUM Control Charts Weakness Strength • Can be constructed as either • Should update the background interwell or intrawell tests. data, and the baseline statistics, every two years • Provides a visual representation • If a release occurs, that data can of compliance data over time and allow for better identification not be used in future of gradual, long term trends. background calculations • Adjustments to certain calculations need to be made if non-detects are >25% U.P. Solid Waste Forum, Marquette MI 4/26/2017 20

• Statistics will only be as good as the data that are used to create them. • Have a data base that has everything , with no changes made to the original results. • What is the purpose of the statistical analyses? • Don’t be afraid to change statistical methods (in fact, you probably should). 4/26/2017 21 U.P. Solid Waste Forum, Marquette MI

References • ASTM. 2010b. Standard Guide for Applying Statistical Methods for Assessment and Corrective Action Environmental Monitoring • Interstate Technology Regulatory Council Programs. D7048-04(2010). West (ITRC) – Web based training course, Conshohocken, PA: ASTM International. Groundwater Statistics for Monitoring and • ASTM. 2012. Standard Guide for Developing Compliance Appropriate Statistical Approaches for • Gibbons, R.D. 1994. Statistical Methods for Groundwater Detection Monitoring Programs. Groundwater Monitoring. New York: John D6312-98(2012)e1. West Conshohocken, PA: Wiley & Sons. ASTM International. • • Dr. Kirk Cameron, MacStat Consulting, Ltd. – USEPA. 2009. "Statistical Analysis of Applied Groundwater Statistics (2 day Groundwater Monitoring Data at RCRA workshop) Facilities." Unified Guidance EPA 530/R-09- 007. Washington DC: United States Environmental Protection Agency. http://www3.epa.gov/epawaste/hazard/correct iveaction/resources/guidance/sitechar/gwstats /unified-guid.pdfguid.pdf. U.P. Solid Waste Forum, Marquette MI 4/26/2017 22

Contact Information Paula Leier-Engelhardt, P.G., C.P.G. Principal Geologist Thank you HydroGeo Solutions LLC 5951 Allen Road | Little Suamico, Wisconsin paula@hydrogeosolutionswi.com 920-737-9811 U.P. Solid Waste Forum, Marquette MI 4/26/2017 23