MAROS: Monitoring and Remediation Optimization System Decision Support Tool and Application



Similar documents
Long-Term Groundwater Monitoring Optimization Taylor Road Landfill Superfund Site Seffner, Hillsborough County, Florida

ROADMAP TO LONG-TERM MONITORING OPTIMIZATION MAY 2005

Risk Management Procedure For The Derivation and Use Of Soil Exposure Point Concentrations For Unrestricted Use Determinations

DRAFT. GROUNDWATER REMEDY COMPLETION STRATEGY: Moving Forward with Completion in Mind

U. S. Army Corps of Engineers Ground Water Extraction System Subsurface Performance Checklist

Gathering and Managing Environmental Quality Data for Petroleum Projects ABSTRACT

COST AND PERFORMANCE REPORT

New Cumberland Army Depot Formerly Used Defense Site (FUDS) Marsh Run Park Site Restoration Community Meeting

Use of Alternate Concentration Limits (ACLs) to Determine Cleanup or Regulatory Levels Under RCRA and CERCLA

Long-Term Monitoring Network Optimization Evaluation. for. Wash King Laundry Superfund Site Lake County, Michigan

STATE OF TENNESSEE DEPARTMENT OF ENVIRONMENT AND CONSERVATION DIVISION OF UNDERGROUND STORAGE TANKS TECHNICAL GUIDANCE DOCUMENT - 008

Risk-Based Decision Making for Site Cleanup

Alternate Concentration Limits/Groundwater Cleanup Levels. Title slide

FMC Environmental Solutions Peroxygen Talk January 2010 Use of Compound Specific Isotope Analysis to Enhance In Situ

DOCUMENTATION OF ENVIRONMENTAL INDICATOR DETERMINATION Interim Final 2/5/99 RCRA Corrective Action Environmental Indicator (EI) RCRIS code (CA725)

ZYMAX NEW FORENSICS ISOTOPE ANALYSIS

MEMORANDUM. Long-Term Monitoring Optimization Review Intel Magnetics/Micro-Storage Corporation Site, Santa Clara, CA

Frequently Asked Questions on the Alberta Tier 1 and Tier 2 Soil and Groundwater Remediation Guidelines. February 2008

Evaluation of Site-Specific Criteria for Determining Potability

Long Term Monitoring Optimization (LTMO) Concepts and tools

Federal Remediation Technologies Roundtable Arlington, VA November 14, 2013 Jim Woolford, Director Office of Superfund Remediation and Technology

Prepared for ENRY2000, Belgrade, Yugoslavia, September 27, 2001

WATER SUPPLY WELL RECEPTOR SURVEY GUIDANCE DOCUMENT

FIVE-YEAR REVIEW REPORT MASON COUNTY LANDFILL SUPERFUND SITE LUDINGTON, MICHIGAN FEBRUARY, 2001

Part II Chapter 9 Chapter 10 Chapter 11 Chapter 12 Chapter 13 Chapter 14 Chapter 15 Part II

EPA Region 3/ Pa Dept of Environmental Protection Streamlining the Process for the One Cleanup Program Under RCRA

Soil Cleanup Goals. Minnesota Department of Agriculture Pesticide and Fertilizer Management Division. Guidance Document 19

Inventory of Performance Monitoring Tools for Subsurface Monitoring of Radionuclide Contamination

APPENDIX N. Data Validation Using Data Descriptors

STATEMENT OF BASIS HYPERGOL SUPPORT BUILDING SWMU 65 NATIONAL AERONAUTICS AND SPACE ADMINISTRATION KENNEDY SPACE CENTER BREVARD COUNTY, FLORIDA

Bioremediation of contaminated soil. Dr. Piyapawn Somsamak Department of Environmental Science Kasetsart University

Analysis of MTBE Groundwater Cleanup Costs Executive Summary

Guidance on Remediation of Petroleum-Contaminated Ground Water By Natural Attenuation

Incorporating Greener Cleanups into Remedy Reviews

HISTORICAL OIL CONTAMINATION TRAVEL DISTANCES IN GROUND WATER AT SENSITIVE GEOLOGICAL SITES IN MAINE

Appendix B: Monitoring Tool Matrices

Guide to the Remediation Report & Fee Submission Form

4.7 HAZARDS AND HAZARDOUS MATERIALS

Review of Groundwater Vulnerability Assessment Methods Unsaturated Zone. Dept. of Earth Sciences University of the Western Cape

ALBERTA TIER 2 SOIL AND GROUNDWATER REMEDIATION GUIDELINES

Alberta Tier 2 Soil and Groundwater Remediation Guidelines

Introduction to Minitab and basic commands. Manipulating data in Minitab Describing data; calculating statistics; transformation.

DOCUMENTATION OF ENVIRONMENTAL INDICATOR DETERMINATION Interim Final 2/5/99 RCRA Corrective Action Environmental Indicator (EI) RCRIS code (CA750)

Presumptive Remedy: Supplemental Bulletin Multi-Phase Extraction (MPE) Technology for VOCs in Soil and Groundwater Quick Reference Fact Sheet

3.3 Data Collection Systems Design Planning Documents Work Plan Field Sampling Plan (FSP)

DRY CLEANING PROGRAM QUALITY ASSURANCE MANAGEMENT PLAN

Overview of the Remedial Action Cost Engineering Requirements (RACER ) Software

Pierre VOC Ground Water Investigation and MTBE Remediation Pilot Study

San Mateo County Environmental Health Characterization and Reuse of Petroleum Hydrocarbon Impacted Soil

Well gauging results LNAPL in Benzol Processing Area

Presented by: Craig Puerta, PE, MBA December 12, 2012

SOIL GAS MODELLING SENSITIVITY ANALYSIS USING DIMENSIONLESS PARAMETERS FOR J&E EQUATION

CHAPTER 4 TYPES OF COST ESTIMATES

RCRA Corrective Action Workshop On Results-Based Project Management: Fact Sheet Series

ANALYSIS OF DNAPL SOURCE DEPLETION COSTS AT 36 FIELD SITES

COMMON CORE STATE STANDARDS FOR

Tim Johnson, Mike Truex, Jason Greenwood, Chris Strickland, Dawn Wellman: Pacific Northwest National Laboratory

Focus on Developing Ground Water Cleanup Standards Under the Model Toxics Control Act

SUCCESSFUL FIELD-SCALE IN SITU THERMAL NAPL REMEDIATION AT THE YOUNG-RAINEY STAR CENTER

Source Water Assessment Report

What Constitutes Environmental Due Diligence?

DOCUMENTATION OF ENVIRONMENTAL INDICATOR DETERMINATION

PROGRAM AT A GLANCE 12

Proxy Simulation of In-Situ Bioremediation System using Artificial Neural Network

New Jersey Department of Environmental Protection (NJDEP) Site Remediation Program INSTRUCTIONS FOR NJDEP ONLINE REMEDIAL INVESTIGATION REPORT SERVICE

Quality Assurance Guidance for Conducting Brownfields Site Assessments

Developing Quality Assurance Project Plans using Data Quality Objectives and other planning tools. Developing QAPPs 5/10/2012 1

Overview of the Remedial Action Cost Engineering Requirements (RACER ) Software

Environmental Accounting Guidelines

N O T E S. Environmental Forensics. Identification of Natural Gas Sources using Geochemical Forensic Tools. Dispute Scenarios

A MODEL OF IN SITU BIOREMEDIATION THAT INCLUDES THE EFFECT OF RATE- LIMITED SORPTION AND BIOAVAILABILITY

STATE OF VERMONT AGENCY OF NATURAL RESOURCES DEPARTMENT OF ENVIRONMENTAL CONSERVATION WASTE MANAGEMENT DIVISION SOLID WASTE MANAGEMENT PROGRAM

Site Cleanup in Connecticut

Vapor Intrusion Pathway: A Practical Guideline

GUIDANCE DOCUMENT FOR DEVELOPMENTS AND SUBDIVISIONS WHERE ONSITE WASTEWATER TREATMENT SYSTEMS ARE PROPOSED

Interim Summary Report for the Treatment Facility D Helipad In-Situ Bioremediation Treatability Test

APPENDIX D RISK ASSESSMENT METHODOLOGY

X X X a) perfect linear correlation b) no correlation c) positive correlation (r = 1) (r = 0) (0 < r < 1)

2. determining that land is not contaminated land and is suitable for any use, and hence can be removed from the CLR or EMR, as relevant.

CURRENT AND FUTURE IN SITU TREATMENT TECHNIQUES FOR THE REMEDIATION OF HAZARDOUS SUBSTANCES IN SOIL, SEDIMENTS, AND GROUNDWATER

Preliminary Estimate Distribution of Exide s Lead Emissions in Soil

Transcription:

MAROS: Monitoring and Remediation Optimization System Decision Support Tool and Application Abstract Julia J. Aziz and Charles J. Newell, Ph.D., P.E., Groundwater Services, Inc., Houston, TX Meng Ling and Hanadi S. Rifai, Ph.D., P.E, University of Houston, Houston, TX Jim Gonzales, Air Force Center for Environmental Excellence, Brooks AFB, TX The Monitoring and Remediation Optimization System (MAROS) software has been developed to provide site managers a strategy for formulating appropriate long-term groundwater monitoring programs. Although the annual cost of long-term monitoring at an individual site may appear relatively small, groundwater monitoring over long time periods creates the potential for a tremendous cost liability of billions of dollars (Air Force Modeling and Monitoring Workshop, August 1997). The MAROS database software optimizes a site-specific monitoring program that is currently tracking the occurrence of contaminant migration in groundwater. MAROS is a decision support tool based on statistical methods applied to site-specific data that accounts for relevant current and historical site data as well as hydrogeologic factors (e.g. seepage velocity) and the location of potential receptors (e.g., wells, discharge points, or property boundaries). Based on this site-specific information, the software suggests an optimization plan for improving the efficiency and effectiveness of the current groundwater well monitoring plan. MAROS includes the application of both temporal and spatial methods through trend analysis (i.e. Mann-Kendall Analysis and/or Linear Regression Analysis and/or Cost Effective Sampling) and well network analysis (i.e. Delaunay Triangulation and/or Moment Analysis) to determine the minimum number of wells and the minimum sampling frequency required for future compliance monitoring at the site. Graphical and spatial visualization tools within the software assist the user in assessing the trend results at each monitoring point. Preliminary monitoring optimization results provide a basis for making more cost effective, scientifically based future long-term monitoring decisions. The MAROS software addresses a variety of groundwater contaminant plumes (e.g., fuels, solvents, and metals) and is designed for continual modification of long-term monitoring plans as the plume or site conditions change. The MAROS software was applied to the site data for 2 chlorinated solvent plumes underlying 3 operable units at McClellan Air Force Base to assess historical trends. These analyses were then consolidated and weighed to obtain synthesized site plume stability information for each plume. The stability information was then used to assess if there were any redundant sampling locations and to determine if the sampling frequency could be reduced. while at the same time meeting future compliance monitoring goals for McClellan AFB. Implementing this revised groundwater monitoring plan could result in significant cost savings for the duration of the monitoring program implementation. The MAROS software is freeware that is available for download at www.gsinet.com. Introduction The need for both active (e.g., pump and treat systems) and passive (e.g., monitored natural attenuation) remediation of affected ground water sites often entails expensive monitoring programs. These long-term monitoring programs, whether applied for process control, performance measurement, or compliance purposes, are dictated by the Resource Conservation and Recovery Act (RCRA), the Comprehensive Environmental Response, Compensation and Liability Act (CERCLA), State Superfund, and the Underground Storage Tank (UST) legislation. In a long-term monitoring program, the scale of the required data collection effort and the time commitment makes its cumulative costs very high. Although the cost of long-term monitoring at individual sites may appear relatively small, ground water monitoring at a large number of sites for long periods of time creates the potential for a tremendous cost liability of over billions of dollars. With the increasing use of risk-based goals and natural attenuation initiatives in recent years as well as the move toward long-term closure upon completion of cleanup activities, the need for better-designed long-term monitoring plans that are both cost-effective and protective of human and ecological health has greatly increased. Developing a methodology for both evaluation and continued re-evaluation of existing long-term monitoring plans can enhance the efficiency of long-term monitoring systems with the potential for substantial cost savings. Few decision-support software tools are available for site managers to quickly and easily assess cost-effective 283

long-term monitoring plans. This paper describes the Monitoring and Remediation Optimization System (MAROS) decision-support software. MAROS is a public domain software and is available online at http://www.afcee.brooks.af.mil/er/rpo.htm or www.gsi-net.com. The MAROS methodology facilitates optimizing long-term monitoring strategies for sites undergoing passive or active remediation. Within the context of this paper, the term "optimization" follows the guidance documents of NFESC, 2000 and AFCEE, 1997, which define optimization as reducing the cost of long-term monitoring while maintaining or increasing the quality and effectiveness. Although this is not the traditional definition of optimization which implies searching the entire solution space for the single best result, in the context of the MAROS methodology an optimal solution is provided together with the parameters within a complicated ground water system which will increase its effectiveness. By applying statistical techniques to existing historical and current site analytical data, as well as by considering hydrogeologic factors and the location of potential receptors, the software suggests an optimal plan along with an analysis of individual monitoring wells. MAROS is designed to address a variety of ground water contaminants (e.g., fuels, solvents, metals), and analyze multiple constituents of concern (COCs) concurrently. MAROS was developed in the Windows based platform using Microsoft Access and Microsoft Excel. Microsoft Access provides a database management backdrop for the software, where automated screens with functions for importing and exporting data, as well as reports were programmed. Microsoft Excel provides features that allow tools for visualization and graphical capabilities to be automated. Microsoft Visual Basic for Applications is used to implement MAROS algorithms and to develop the user interface. All software screens are simple and easy to use and do not require any knowledge of Microsoft Access or Microsoft Excel. MAROS Methodology In MAROS, both spatial and temporal analysis tools are available for evaluating long-term monitoring plans: temporal trend analysis results in plume stability information; and sampling optimization methodology is based on spatial and temporal redundancy reduction methods. The interpretive trend analysis approach assesses the general monitoring category by considering individual well concentration trend, overall plume stability, hydrogeologic factors (e.g., seepage velocity), and the location of potential receptors (e.g., property boundaries or drinking water wells). The sampling optimization methodology consists of a sampling location analysis through the Delaunay method and a sampling frequency analysis through the Modified CES method. The Delaunay method is designed to minimize monitoring locations and the Modified CES method is designed to minimize the frequency of sampling. MAROS Structure The software flow diagram of MAROS is shown in Figure 1. Data Management provides data import and export functions. Site Details are entered pertaining to sampling events, well classification (source or tail well), and the constituents of concern (COCs). Data Consolidations transforms the raw monitoring records to useable quantitative data for further analysis. The trend analysis approach is implemented to gain information about each well s concentration over time, as well as an overall plume trend in both the source and tail regions. Sampling Optimization determines the detailed location and frequency of sampling for a specific site. The MAROS software can be utilized in a step-by-step manner, with each progressive step yielding information that can answer site-specific compliance monitoring questions. At each step in the software, reports or graphs of results can be readily generated. 284

Figure 1. The software flow diagram of MAROS. Data Management In MAROS, ground water monitoring data can be imported from simple database-format Microsoft Excel spreadsheets, ERPIMS (Environmental Resources Program Information Management System, developed by AFCEE) files, or previously created MAROS database archive files, or the data can be entered manually. Compliance monitoring data interpretation in MAROS is based on historical ground water monitoring data from a consistent set of wells over a series of sampling events. Statistical validity of the concentration trend analysis requires constraints on the minimum data input from at least four wells (ASTM 1998) in which COCs have been detected. Individual sampling locations need to include data from at least six most-recent sampling events. To ensure a meaningful comparison of COC concentrations over time and space, both data quality and data quantity need to be considered. Prior to statistical analysis, the user can consolidate irregularly sampled data or smooth data that might result from seasonal fluctuations or changes in site conditions. Imported ground water monitoring data and the site-specific information entered in Site Details can be archived and exported as MAROS archive files. These archive files can be appended as new monitoring data becomes available, resulting in a dynamic long-term monitoring database that reflects the changing conditions at the site (i.e. biodegradation, individual well compliance attainment, completion of remediation phase, etc.). Site Details Information needed for the MAROS analysis includes site-specific parameters such as seepage velocity and current plume length. Part of the trend analysis methodology applied in MAROS focuses on where the 285

monitoring well is located, therefore the user needs to divide site wells into two different zones: the source zone and the tail zone. The source zone includes areas with non-aqueous phase liquids (NAPLs), contaminated vadose zone soils, and areas where aqueous-phase contaminant releases have been introduced into groundwater. The source zone generally contains locations with historical high ground water concentrations of the COCs. The tail zone is usually the area downgradient of the contaminant source zone. Although this classification is a simplification of the well location, this broadness makes the user aware, on an individual well basis, that the concentration trend results can have a different interpretation depending on the well location in and around the plume. It is up to the user to make further interpretation of the trend results, depending on the type of well analyzed (e.g., remediation well, leading plume edge well, or monitoring well). MAROS allows the analysis of up to 5 COCs concurrently and users can pick COCs from a list of compounds existing in the monitoring data, or select COCs based on recommendations provided in MAROS based on toxicity, prevalence, and mobility of compounds. Sites with more than 5 COCs can be analyzed with multiple MAROS runs. In general, the MAROS methodology applies to 2-D aquifers that have relatively simple site hydrogeology. However, for a multi-aquifer (3-D) system, the user could apply the statistical analysis layer-bylayer. Data Consolidation Typically long-term monitoring raw data have been measured irregularly in time or contain many nondetects, trace level results, and duplicates. Therefore, before the data can be further analyzed, raw data are filtered, consolidated, transformed, and possibly smoothed to allow for a consistent dataset that meets the minimum data requirements for statistical analysis mentioned previously. MAROS allows users to specify the period of interest in which data will be consolidated (i.e., monthly, bimonthly, quarterly, semi-annual, yearly, or biennial basis). In computing the representative value when consolidating, one of four statistics can be used: median, geometric mean, mean, and maximum. Nondetects can be transformed to one half the reporting or method detection limit (DL), the DL, or a fraction of the DL. Trace level results can be represented by their actual values, one half of the DL, the DL, or a fraction of their actual values. Duplicates are reduced in MAROS by one of three ways: assigning the average, maximum, or first value. Consolidation of data is often a very tedious, time consuming and expensive process, however, MAROS automates this process. The reduced data for each COC and each well can be viewed as a time series in a graphical form on a linear or semi-log plot generated by the software. Temporal Trend Analysis Within the MAROS software there are historical data analyses that support a conclusion about plume stability (e.g., increasing plume, etc.) through statistical trend analysis of historical monitoring data. Plume stability results are assessed from time-series concentration data with the application of two statistical tools: Mann- Kendall analysis and linear regression analysis. The two methods are used to estimate the concentration trend for each well and each COC based on a statistical trend analysis of concentrations versus time at each well. Nonparametric tests such as the Mann-Kendall test for trend are suitable for analyzing data that do not follow a normal distribution. Nonparametric methods focus on the location of the probability distribution of the sampled population, rather than specific parameters of the population. The outcome of the test is not determined by the overall magnitude of the data points but depends on the ranking of individual data points. Data distribution assumptions are not necessary for nonparametric tests. The Mann-Kendall test has no distributional assumptions and irregularly spaced measurement periods are permitted. This approach has advantages in cases where outliers in the data would produce biased estimates of the least squares estimated slope. Mann-Kendall Trend Analysis The Mann-Kendall test is a non-parametric statistical procedure that is well suited for analyzing trends in data over time (Gilbert, 1987). The Mann-Kendall test can be viewed as a nonparametric test for zero slope of the first-order regression of time-ordered concentration data versus time. This procedure does not require knowing the statistical distribution of the data and can be used with data sets that include irregular sampling 286

intervals and missing data. The Mann-Kendall test can also be used with data reported as trace or less than the method DL because it uses only the relative magnitudes of the data rather than their measured values. This test has advantages in cases where outliers in the data would produce biased estimates of the least squares estimated slope. The Mann-Kendall analysis determines the concentration trend by considering three factors: the Mann- Kendall statistic from the Mann-Kendall test (Gilbert 1987), the confidence in the trend, and the coefficient of variation of the concentration data. The Mann-Kendall test is designed for analyzing a single groundwater constituent; multiple constituents are analyzed separately. The Mann-Kendall statistic (S) measures the trend in the data with positive values indicating an increase in constituent concentrations over time and negative values indicating a decrease in constituent concentrations over time. The Mann-Kendall concentration trend is classified into six categories: Decreasing (D), Probably Decreasing (PD), Stable (S), No Trend (NT), Probably Increasing (PI), and Increasing (I) depending on statistical indicators (see Table 1, for a more detailed explanation see Aziz, et. al., 2002). These temporal trend estimates are then analyzed in MAROS Analysis to identify plume stability, which is then classified into the similar categories as the concentration trend. The raw imported data must meet minimum data requirements as to the frequency of sampling, duration of the sampling intervals for trend analysis, and sampling density for the site as well as the quality of the measurements. Table 1 The Mann-Kendall Analysis Decision Matrix (Aziz, et. al., 2002) Mann-Kendall Statistic Confidence in the Trend Concentration Trend S > 0 > 95% Increasing S > 0 90-95% Probably Increasing S > 0 < 90% No Trend S 0 < 90% and COV 1 No Trend S 0 < 90% and COV < 1 Stable S < 0 90-95% Probably Decreasing S < 0 > 95% Decreasing Linear Regression Analysis Linear Regression is a parametric statistical procedure that is typically used for analyzing trends in data over time. Using this type of analysis, a higher degree of scatter simply corresponds to a wider confidence interval about the average log-slope. Assuming the sign (i.e., positive or negative) of the estimated log-slope is correct, a level of confidence that the slope is not zero can be easily determined. Thus, despite a poor goodness of fit, the overall trend in the data may still be ascertained, where low levels of confidence correspond to Stable or No Trend conditions (depending on the degree of scatter) and higher levels of confidence indicate the stronger likelihood of a trend. The coefficient of variation, defined as the standard deviation divided by the average, is used as a secondary measure of scatter to distinguish between Stable or No Trend conditions for negative slopes. The Linear Regression Analysis is designed for analyzing a single groundwater constituent; multiple constituents are analyzed separately, (up to five COCs simultaneously). For this evaluation, a decision matrix developed by Groundwater Services, Inc. was used to determine the Concentration Trend category (plume stability) for each well. 287

Table 2 The Linear Regression Analysis Decision Matrix (Aziz, et. al., 2002) Confidence in the Log-slope Trend Positive Negative < 90% No Trend COV < 1 Stable COV > 1 No Trend 90-95% Probably Increasing Probably Decreasing > 95% Increasing Decreasing Parametric tests such as first-order regression analysis assume the data follows a normal distribution, allowing results to be affected by outliers in the data in some cases. However, parametric methods result in more accurate trend assessments where there is a normal distribution of the residuals. Therefore, when the data is normally distributed the non-parametric method, Mann-Kendall test, may not be as powerful. Both tests are utilized in the MAROS software. For an adequately delineated plume, a stable or shrinking condition can be identified by a stable or decreasing concentration trend over time. The MAROS trend analysis provides an overall plume condition for each well and Constituent of Concern (COC) based on a statistical trend analysis of concentrations at each well over time. Under optimal conditions, the natural attenuation of COCs at any site is expected to approximate a first-order exponential decay for compliance monitoring groundwater data. With actual site measurements, apparent concentration trends may often be obscured by data scatter arising from non-ideal hydrogeologic conditions, sampling and analysis. However, even though the scatter may yield a poor goodness of fit (typically characterized by a low correlation coefficient, e.g., R 2 << 1) for the first-order relationship, parametric and nonparametric methods can be utilized to obtain confidence intervals on the estimated first-order coefficient (the slope obtained from the log-transformed data). Sampling Optimization In order to provide more specific and directed long-term monitoring plans, more rigorous sampling optimization methods were developed and included in MAROS for sampling location and sampling frequency analyses. The Delaunay method for sampling location analysis and the Modified CES method for frequency analysis are described in the next two sections. Sampling Location determination uses the Delaunay triagulation method to determine the significance of the current sampling locations relative to the overall monitoring network. The Delaunay method converts data from multiple wells to a single value representing the area concentration, which is then compared to the average plume concentration. A slope factor (SF) is calculated for each well to indicate the significance of this well in the system (i.e. how removing a well changes the average concentration.) The Sampling Location optimization process is performed in a stepwise fashion. Step one involves assessing the significance of the well in the system, if a well has a small SF, and the Area and Average Concentration after eliminating the well is close to the original values, the well can be removed from the monitoring network. Step two involves the significance of the well information. If one well has a small SF, it may or may not be eliminated depending on whether the information loss is significant. If the information loss is not significant, the well can be eliminated from the monitoring network and the process of optimization continues with fewer wells. However if the well information loss is significant then the optimization terminates. This sampling optimization process will allow the user to assess redundant wells that will not incur significant information loss on a constituent by constituent basis for individual sampling events. Since the Delaunay method is performed for each COC, results are given separately for each COC. Integrated results based on all COCs are also provided, in which a sampling location is marked as eliminated only when it is eliminated for all COC 288

analyses.the Delaunay method is available in both graphical and non-graphical form within the MAROS software. Sampling Frequency determination applies a Modified Cost Effective Sampling (MCES) method (Ridley, et. al, 1998) to compare current and long-term COC plume stability trends based on time series sampling data to assess the feasibility of reducing the sampling frequency. Linear Regression and Mann-Kendall analysis are adopted for estimating COC trends. The variability of the sequential sampling data has been accounted for by the Mann-Kendall analysis. The result of this step is a suggested sampling frequency based on recent sampling data trends and overall sampling data trends. The Modified CES method was developed in the MAROS software to determine the optimal sampling frequency for each sampling location in the monitoring network. The Modified Cost Effective Sampling is a methodology for estimating the lowest-frequency sampling schedule for a given groundwater monitoring location and can still provide needed information for regulatory and remedial decision-making. It aims to reduce sampling frequency based on the analysis of time series sampling data in each well, considering both current trends and long term trends of the concentration of contaminants. In order to estimate the least frequent sampling schedule for a monitoring location that still provides enough information for regulatory and remedial decision-making, MCES employs three steps to determine the sampling frequency. The first step involves analyzing frequency based on recent trends. A preliminary location sampling frequency (PLSF) is determined based on the trends determined by rates of change from linear regression and Mann-Kendall analysis of the most recent monitoring data. The variability of the sequential sampling data has been accounted for by the Mann-Kendall analysis. The PLSF is then adjusted based on overall trends. If the long-term history of change is significantly greater than the recent trend, the frequency may be reduced by one level. Otherwise, no change could be made. The final step in the analysis involves reducing frequency based on risk. Since not all compounds in the target being assessed are equally harmful, frequency is reduced by one level if recent maximum concentration for compound of high risk is less than 1/2 of the Maximum Concentration Limit (MCL). The result of applying this method is a suggested sampling frequency based on recent sampling data trends and overall sampling data trends. The sampling optimization methodology described above as wells as the temporal trend and plume stability analysis have been applied to a number of sites around the country. The following section presents one such application at a military facility to demonstrate the benefits from applying this integrated methodology. Site Application: McClellan Air Force Base The MAROS methodology was applied to the Zone A and Zone B aquifers in Operable Units A and B/C at the McClellan Air Force Base to evaluate the current long-term monitoring network in use at the site. McClellan AFB is located in the Sacramento Valley, northeast of Sacramento, California. The shallow geologic units under the base consist primarily of alluvial and fluvial sediments eroded from the Sierra Nevada. The depth to groundwater is typically 100 feet below ground surface (bgs). The ground water flow direction is predominantly toward the south-southwest and the representative ground water seepage velocity is 0.3 m/day. The base was established in 1936 as an aircraft repair depot and supply base. The base has been divided into eight operable units (OUs, A, B, etc.) when the Interagency Agreement governing the environmental cleanup program at the McClellan AFB was signed in 1988. Groundwater underlying the base has been contaminated by chlorinated solvents resulting from past handling and disposal practices related to base operations. The four most significant contaminants are Trichloroethene (TCE); Tetrachloethene (PCE), 1,2-Dichloroethane and cis- 1,2-Dichloroethene were cited in the Interim Record of Decision (IROD) and are used to evaluate progress toward IROD goals (Radian, 1999). Groundwater contamination was first discovered in wells on and adjacent to the base in 1979. Subsequent investigations were conducted to define the type, magnitude, and extent of contamination. As a result of these investigations, five groundwater monitoring zones (A through E, increasing with depth) have been defined based on geologic and hydrologic characteristics. Two large groundwater plumes and several smaller plumes have been identified. The larger plumes (OU A and OU B/C) are almost entirely defined by the extent of TCE contamination and exist only at depths extending to the A and B monitoring zones (zone thickness 20 and 50 feet respectively). 289

The TCE Plume OU A is 5000 feet long, between 2,000 to 3,000 feet wide and 20 to 50 feet thick. The TCE Plume OU B/C is 5800 feet long, between 2,000 to 3,800 feet wide and 20 to 50 feet thick. The ground water long-term monitoring plan consists of remediation performance monitoring and compliance monitoring in order to verify plume containment and monitor plume reduction to verify progress toward achieving cleanup goals. Groundwater monitoring of TCE at the site has fluctuated between quarterly and semiannual sampling for the years 1996 to 2000. By the 4 th Quarter 2000, 20 sampling events had been carried out at the site. However, not all wells were included in each sampling event and consequently some of the wells did not have sufficient data to be analyzed by trend analysis (a minimum of 2 years data is required). The irregular frequency of sampling stems from different times that extraction systems and long-term monitoring programs were implemented at each operating unit (OU) or monitoring wells have been installed over time as the plume has changed. A two-dimensional plume analysis was assumed adequate. Two TCE plumes have been analyzed with the MAROS software. Source wells have been selected based on the criteria of monitoring/extraction wells with historical concentrations in excess of 100*MCL for TCE. Between 1987 and 1995, six groundwater treatment systems were installed and operated to treat contaminated groundwater at McClellan AFB. The remediation system and monitoring for plume OU A consists of 3 wellhead treatments systems connected to 7 extraction wells and approximately 60 monitoring wells in and around the plume. The remediation system and monitoring for plume OU B/C and OU C consists of an extraction system with 11 extraction wells and approximately 60 monitoring wells. In applying MAROS to develop a monitoring strategy for the TCE plume, trend results from the TCE Mann- Kendall analysis and the linear regression analysis for both source and tail wells were considered. Both Mann- Kendall and Linear Regression analyses were run on both zones of both plumes and gave similar trend estimates for each well. In evaluating plume stability, the primary lines of evidence results and all monitoring wells were assigned Medium weights, assuming equal importance in the overall analysis. For both plumes and both zones, the source stability and the tail stability were determined to be Stable, Decreasing, or Probably Decreasing overall, resulting in a Moderate monitoring system category. The results from applying the trend analysis methods in MAROS to the TCE plume are summarized in Table 1. Plume OU A: 7 out of 13 source wells and 26 out of 47 tail wells have a Stable, Decreasing or Probably Decreasing trend for Plume OU A, the rest of the wells show Increasing or No Trend. Plume OU B/C:4 out of 5 source wells and 40 out of 62 tail wells have a Stable, Decreasing or Probably Decreasing trend for Plume OU A, the rest of the wells show Increasing or No Trend. However, monitoring wells and extraction wells are present in both well networks and need to be treated differently for the purpose of individual trend analysis interpretation primarily due to the different course of action possible for the two types of wells. For monitoring wells, strongly decreasing concentration trends may lead the site manager to decrease their monitoring frequency. Conversely, strongly decreasing concentration trends in extraction wells may indicate ineffective or near-asymptotic contamination extraction, which may in turn lead to either the shutting down of the well or a drastic change in the extraction scheme. Other reasons favoring the separation of these two types of wells in the trend analysis interpretation is the fact that they produce very different types of samples. On average, the extraction wells at McClellan possess screens that are twice as large and diameters that are 50% greater than those of the average monitoring well. Therefore, the potential for the dilution of extraction well samples is far greater than either monitoring well or piezometer samples. 290

The results from applying the sampling optimization methods in MAROS to the TCE plume are summarized in Table 1. In the spatial analysis application: Plume OU A: 12 monitoring wells were identified as candidates for elimination from the well sampling network by analyzing the between 8 and 12 recent sampling events, that is, there are 12 possible redundant wells among the 60 wells being analyzed. Plume OU B/C: 16 monitoring wells were identified as candidates for elimination from the well sampling network by analyzing the between 8 and 12 recent sampling events, that is, there are 16 possible redundant wells among the 62 wells being analyzed. The Slope Factor threshold values used for elimination criteria are 0.10 and 0.01 for nodes inside and on the edge of the triangulation domain, respectively. No wells at the leading edge of the plume were eliminated. The resulting values of the Concentration Ratio and the Area Ratio were both within the range of 0.95 ~ 1.00, meaning that more than 95% spatial information was kept. Temporal analysis over all 20 sampling events resulted in significant reduction in the sampling frequency of monitoring wells. This analysis indicates there is a potential to sample 50% of the wells less frequently than the current quarterly sampling frequency. The recommended long-term monitoring strategy results in considerable reduction in sampling costs and allows site managers to develop a better understanding of plume behavior with time. Table 1. The following table summarizes preliminary results of applying MAROS to McClellan AFB TCE plumes. Plume MAROS Trend analysis Sampling optimization analysis Name Plume Source Stability Plume Tail Stability Number of Well Candiates for elimination Frequency Analysis of Wells OU-A, Zone A OU-A, Zone B OU-B/C, Zone A OU-B/C, Zone B Trend Results (Number of Wells) S S S+D+PD=26 I+PI=5 NT=9 PD NT S+D+PD=7 I+PI=6 NT=6 S S S+D+PD=35 I+PI=12 NT=7 D S S+D+PD=9 I+PI=2 NT=1 Number of wells Analyze d 41 10 19 2 54 14 12 2 Note: Events used analyzed Q1 1997 ~ Q4 2000. Note: Events analyzed are Q1 1996 ~ Q4 2000. Note: Events analyzed are Q3 1999 ~ Q4 2000. Note: Events analyzed are Q1 1995 ~ Q4 2000. Annual: 26 Quarterly: 15 Note: Events Analyzed Q1 1997 ~ Q4 2000. Annual: 10 Quarterly: 9 Note: Events Analyzed Q1 1997 ~ Q4 2000. Annual: 29 Quarterly: 25 Note Events analyzed Q1 1996 ~ Q4 2000. Annual: 7 Quarterly: 5 Note: Events analyzed Q1 1997 ~ Q4 2000. Note: Decreasing (D), Probably Decreasing (PD), Stable (S), No Trend (NT), Probably Increasing (PI), and Increasing (I); The following parameters are used in performing the sampling location analysis: Inside SF threshold is 0.10, Hull SF threshold is 0.01, CR and AR thresholds are both 0.95. Conclusions In recent years, the high cost of long-term monitoring as part of active or passive remediation of affected ground water has made the design of efficient and effective ground water monitoring plans a pressing concern. Periodically updating and revising long-term monitoring programs with changing conditions at the site can mean considerable savings in site monitoring costs. The MAROS decision-support software presented in this 291

paper assists in revising existing long-term monitoring plans based on the historical and current monitoring data and plume behavior over time. MAROS is an easy-to-use software developed to provide site managers with a strategy to optimize an existing ground water long-term monitoring program. A temporal trend analysis approach and a sampling optimization methodology are implemented in MAROS to determine the minimum number of sampling locations, optimal sampling frequency, and duration required for future compliance monitoring at the site. The temporal trend analysis approach provides individual concentration trend results as well as general monitoring suggestions based on overall plume stability and site information. The sampling optimization methodology, consisting of the Delaunay method and the Modified CES method, provides detailed recommendations for sampling locations and frequency that could be used for future monitoring at the site. MAROS addresses a variety of groundwater contaminants (e.g., fuels, solvents, and metals) and can handle up to 5 COCs concurrently. MAROS is designed to keep track of a long-term monitoring plan and modify it as the plume or site condition changes over time.the current version of MAROS (Version 1.0) primarily assesses redundancy reduction in existing well networks. However, there are other issues that could impact the overall efficiency and adequacy of long-term ground water monitoring networks. MAROS Version 2.0 is currently under development and will include an evaluation of the sufficiency of an existing monitoring program in delineating the plume through the application of statistical power analysis. MAROS Version 2.0 will utilize power analysis to evaluate statistical power, sample size, and cleanup status associated with monitoring plans. Additionally, expanded visualization tools as well as a moment analysis module are being developed to enhance data managing and processing. References AFCEE, 1997, AFCEE Long-Term Monitoring Optimization Guide, http://www.afcee.brooks.af.mil. Aziz, J, Ling, M., Newell, C., Rifai, H., and Gonzales, J., 2002, MAROS: a Decision Support System for Optimizing Monitoring Plans, Ground Water, In Press. Gilbert, R. O., 1987, Statistical Methods for Environmental Pollution Monitoring, Van Nostrand Reinhold, New York, NY, ISBN 0-442- 23050-8. NFESC. 2000. Guide to Optimal Groundwater Monitoring - Interim Final. Naval Facilities Engineering Service Center, Port Hueneme, CA. Radian International, 1999. Installation Restoration Program Five Year Review Report for McClellan AFB/EM, USAF Contract No. F04699-93-D0018, Delivery Order No. 8062. Ridley, M. et al, 1998, Cost-Effective Sampling of Groundwater Monitoring Wells, the Regents of UC/LLNL, Lawrence Livermore National Laboratory. Biographical Sketches: Julia J. Aziz is an environmental engineer with Groundwater Services, Inc. She has an M.S. from Rice University and a B.A. in Geology from The Johns Hopkins University. Ms. Aziz has extensive experience with the design and programming of geologic and environmental database software as well as GIS applications for commercial and government clients. She has coded many of the software utilities and designed the interface for the MAROS database software. Currently, she is the Project Manager for an AFCEE-funded project to incorporate additional LTM tools in the MAROS software and apply MAROS at several Air Force Bases. J. Aziz, Groundwater Services, Inc., 2211 Norfolk, Suite 1000, Houston, TX 77098 Phone: (713) 522-6300, FAX: (713) 522-8010, e-mail: jjaziz@gsi-net.com. Dr. Charles J. Newell is a Vice President with Groundwater Services, Inc. (GSI) located in Houston. He received a Ph.D. in Environmental Engineering from Rice University and is a member of the American Academy of Environmental Engineers. He is currently an adjunct professor in the Department of Environmental Science and Engineering at Rice. He has co-authored two groundwater books, "Groundwater Contamination: Transport and Remediation" published by Prentice-Hall and "Natural Attenuation of Fuels and Chlorinated Solvents" published by John Wiley & Sons, and is the lead author of three EPA groundwater 292

publications. Dr. Newell developed the BIOSCREEN natural attenuation model for the U.S. Air Force, which is now the most widely distributed groundwater model from EPA's Center for Subsurface Modeling Support and directed development of the BIOCHLOR model and MAROS Monitoring Optimization system. C. Newell, Groundwater Services, Inc., 2211 Norfolk, Suite 1000, Houston, TX 77098, Phone: (713) 522-6300, e-mail: cjnewell@gsi-net.com. Meng Ling is a Ph.D. candidate at the University of Houston. He studies in the Department of Civil and Environmental Engineering under the direction of Dr. Hanadi S. Rifai. His research focuses on groundwater monitoring optimization and groundwater modeling. He has programmed the Sampling Optimization part for the MAROS software.. M. Ling, University of Houston, Dept. of Civil and Environmental Engineering, Houston, Texas 77204-4791. Phone: (713) 743-0751 e-mail: meng.ling@mail.uh.edu. Dr. Hanadi S. Rifai is an Associate Professor at the University of Houston in Houston, Texas. She teaches courses in Engineering Design, Computers in Engineering, Hazardous Waste Management and Risk Assessment and Geographical Information Systems. She is a co-principal Investigator for the AFCEE MAROS monitoring optimization project. Dr. Rifai s research interests are focused on the solution of environmental problems in surface waters and ground water systems. She has completed nationally recognized research on the natural attenuation of contaminants in the environment, modeling biodegradation and bioremediation systems and the application of Geographical Information Systems. H.S. Rifai, University of Houston, Dept. of Civil and Environmental Engineering, Houston, Texas 77204-4791. Phone: (713) 743-4271 Fax: (713) 743-4260 e- mail: Rifai@UH.edu James Gonzales is an Environmental Engineer at the Air Force Center for Environmental Excellence (AFCEE) in the Technology Transfer Division where he manages the bioremediation section. The Division is charged with identifying and field testing new and innovative environmental technologies for use by the U.S. Air Force. Mr. Gonzales holds bachelor degrees in Microbiology and Chemistry from Texas Tech University, and an MS in Chemical Engineering and MBA from the Texas A&M University System. J. Gonzales, AFCEE/ERT, 3201 Sidney Brooks, Bldg. 532, Brooks AFB, TX 78235, Phone (210) 536-4324, e-mail: james.gonzales@hqafcee.brooks.af.mil. 293