Use of Historical Pump-and-Treat Data to Enhance SiteCharacterization and Remediation Performance Assessment

Mark L. Brusseau

Received: 11 June 2013 /Accepted: 4 September 2013 /Published online: 19 September 2013# Springer Science+Business Media Dordrecht 2013

Abstract Groundwater withdrawal and contaminantconcentration data are routinely collected for pump-and-treat operations conducted at hazardous wastesites. These data sets can be mined to produce a wealthof information to support enhanced site characteriza-tion, optimization of remedial system operations, andimproved decision making regarding long-term sitemanagement and closure. Methods that may be usedto analyze and interpret pump-and-treat data to pro-duce such assessments are presented, along with a briefillustration of their application to a site. The resultspresented herein illustrate that comprehensive analysisof pump-and-treat data is a powerful, cost-effectivemethod for providing higher-resolution, value-addedcharacterization of contaminated sites.

Keywords DNAPL .Mass flux . Source depletion

1 Introduction

Pump and treat is currently the primary method used tocontain and treat groundwater contaminant plumes atmany hazardous waste sites. Groundwater withdrawaland contaminant concentration data are routinely col-lected under regulatory requirement for these pump-

and-treat operations. However, these data are rarelyused for purposes other than to monitor the mass ofcontaminant removed. These data sets constitute asource that can be mined to provide additional infor-mation to enhance site characterization activities andremediation performance assessments (Brusseau et al.2007, 2011a, b).

Analysis of historical pump-and-treat data hasthe potential, for example, to provide the followinginformation:

1. Estimates of initial contaminant mass2. Time-continuous measurements of contaminant

mass discharge3. Time-continuous measurements of magnitudes

and rates of mass removal4. Characterization of the relationship between re-

ductions in contaminant mass discharge and reduc-tions in contaminant mass

5. Delineation of source-zone architecture6. Delineation of contaminant mass-removal conditions7. Assessment of contaminant persistence8. Identification of specific mass-transfer processes

and other factors influencing mass removal9. Assessment of the impact of source-zone remedial

actions on overall risk reduction

The information obtained frommining of the data canbe used to update the site conceptual model, to revise thedesign and operation of the remediation systems, and tosupport decision making concerning remedy modifica-tion, long-term site management, and closure. The ob-jective of this brief communication is to illustrate the

Water Air Soil Pollut (2013) 224:1741DOI 10.1007/s11270-013-1741-8

M. L. Brusseau (*)School of Earth and Environmental Sciences,University of Arizona,429 Shantz, Tucson, AZ 85721, USAe-mail: brusseau@email.arizona.edu

types of data sets and associated evaluations that can beobtained frommining of pump-and-treat operations data.

2 Method Description

The basic approach of the characterization methodinvolves the following components:

1. Tabulation and quality assurance evaluation of rawgroundwater withdrawal and contaminant concen-tration data. This step entails collection of the rawdata from the relevant sources and evaluation ofvarious data quality aspects. For example, what isthe frequency of data collection? For systems withmultiple extraction wells, are data collected forindividual wells, or only for the composite inflowinto the treatment system? Periods of system shut-down should be noted and their impact on calcu-lations accounted for in the analyses.

2. Calculation of contaminant mass discharge as afunction of operational time. Contaminant mass-discharge values are calculated for each measure-ment period as the product of the extraction wellpumpage and mean contaminant concentration.These values are equivalent to the magnitude ofcontaminant mass removed with the pump-and-treat system for the given measurement period.The data are typically converted to standard unitssuch as kilogram per day.

3. Integration of the temporal mass-discharge data todetermine time-continuous measurements of mag-nitudes and rates of mass removal. Stepwise inte-gration of the contaminant mass-discharge data setprovides magnitudes of contaminant mass re-moved as a function of operational time. Plots ofcumulative contaminant mass removed are oftenconstructed to help visualize mass-removal behav-ior. The contaminant mass-discharge data also pro-vide the equivalent of mass-removal rates.

4. Application of mathematical models or functionsto estimate the mass of contaminant that was pres-ent in the treatment zone at the start of remediation.The total contaminant mass initially present in thesource area at a site is a critical variable that isunknown for most field sites. The standard methodused to estimate initial mass, based on collectionand analysis of sediment core samples, is expen-sive and typically influenced by a large degree of

uncertainty. An alternative approach is based onfitting a source-depletion function to temporal con-centration or contaminant mass-discharge data.

In general, a mechanistic-based reactive trans-port model can be calibrated to historical concen-tration data to solve the inverse problem for initialmass. However, the use of advanced transportmodels for field sites is typically constrained by alack of information needed to parameterize themodel. In lieu of this approach, simplifiedsource-depletion functions can be fit to measureddata to estimate initial mass. For example, simpli-fied functions have been fit to temporal concentra-tion data collected from monitoring wells locatedwithin contaminant plumes to provide estimates ofsource mass (Butcher and Gauthier 1994; Basuet al. 2009). This approach has recently been ap-plied to contaminant mass-discharge data obtainedfrom analysis of pump-and-treat operation data(Brusseau et al. 2013).

5. Use of calculated initial mass and time-continuousmass-removal data to characterize the relationshipbetween reductions in contaminant mass discharge(CMDR) and reductions in contaminant mass(MR). The CMDR–MR relationship is a definingcharacteristic of system behavior and is mediatedby system properties and conditions such as per-meability distribution, contaminant distribution,and mass-transfer processes. This relationship isuseful for delineating mass-removal conditions ofthe system and thus has the potential to be a pow-erful tool for assessing the performance of reme-diation operations. Several prior CMDR–MR pro-files have been reported based on laboratory ex-periments and mathematical modeling, while onlya few have been reported for field sites (Brusseauet al. 2007, 2013; DiFilippo and Brusseau 2008).

6. Analysis of the processed data to evaluate contam-inant mass-removal conditions, source-zone archi-tecture, contaminant persistence, specific mass-transfer processes and other factors influencingmass removal, and the impact of source-zone re-medial actions on overall risk reduction. This stepemploys the data produced in the prior steps, incombination with other site information such ascontained within the site conceptual model, andcan be enhanced by the application of mathemati-cal modeling. The type of model used can rangefrom simple ones that focus on a single transport

1741, Page 2 of 6 Water Air Soil Pollut (2013) 224:1741

process, such as diffusion models, or that employ alumped mass-transfer term (e.g., source-depletionmodels) to complex transport and fate models thatattempt to account explicitly for each relevant fac-tor and process. The specific model used will de-pend on the availability of information and datarequired for model input, as well as the objectivesof the effort. An example application of a complexmodel to a pump-and-treat system was reported byZhang and Brusseau (1999), who characterized therelative impacts of plume-scale back diffusion,plume-scale sorption/desorption, and dissolutionof organic liquid trapped in the source zones onperformance of a pump-and-treat system for a sitein Tucson, AZ, USA. It was shown that the asymp-totic conditions observed for mass removal werecaused primarily by the uncontrolled source zones.The results of additional modeling indicated thatthe plume would persist for many decades evenwith isolation or remediation of the source zones,primarily due to back diffusion of contaminantassociated with the lower-permeability units(Brusseau et al. 2007).

While the use of advanced transport and fatemodels provides the most robust level of assess-ment, their application to field sites is typicallyconstrained by a lack of information needed toparameterize the model. Thus, simplified ap-proaches are often used. One example of a simpli-fied approach is the use of source-depletion func-tions, which have recently received increased at-tention. This approach is often used when thespecific processes and/or site conditions influenc-ing contaminant transport are incompletely char-acterized, and thus, the use of process-specificmodels is problematic. An example of such a func-tion is the first-order, exponential function, givenas CMDt/CMD0 = exp(−kt) (e.g., Zhu and Sykes2004; Falta et al. 2005), where CMD is contami-nant mass discharge, t is time, k is the depletioncoefficient, and subscripts t and 0 represent mea-sured and initial values, respectively.

3 Case Study

Data collected from a federal Superfund site (i.e., listed onthe Environmental Protection Agency, EPA, National

Priorities List) located in Arizona are used herein to illus-trate the methods. The site is contaminated by chlorinated-solvent compounds (primarily trichloroethene), and a largegroundwater contaminant plume is present. A pump-and-treat system has been in operation at the site for approxi-mately 18 years. The raw groundwater withdrawal andcontaminant concentration data were graciously providedby the EPA project managers and subcontractors workingwith the EPA and the responsible parties.

The aquifer at the site comprises sand and gravelalluvium, with an ~20-m-thick silty clay unit in themiddle that represents roughly less than one quarter ofthe total treatment-zone thickness. Solvent disposaloccurred via injection into shallow dry wells. Thepump-and-treat well field is designed such that thereare no extraction wells located within the source area.A single extraction well is located approximately100 m downgradient from the source zone, with allother extraction wells distributed further within theplume. These wells are grouped in two sets: one inrelative close proximity to the source extraction welland the other at the downgradient margin of the plume,with the two sets approximately 1,500 m apart. Treatedgroundwater is reinjected into wells located farupgradient of the source zone.

4 Results and Discussion

The historical contaminant mass discharge (CMD) de-termined from analysis of the pump-and-treat data ispresented in Fig. 1. The initial CMD was approximate-ly 10 kg/day. This value is quite large in comparison tothe range of values reported in a recent summary(ITRC 2010). This is reflective of the significant im-pact of the highly contaminated source zone, owing tothe large quantities of solvent disposed of therein.

Interestingly, asymptotic behavior is observed forCMD in the later stage of operation. This may beattributed to the impact of constraints to mass transfer,such as poorly accessible contaminant mass associatedwith solvent in source zones as well as mass stored inlower-permeability units. Other factors, such as theimpact of well-field hydraulics, may also influencethe observed behavior.

To illustrate the application of a simplified transportmodel, a model employing a dual-porosity conceptu-alization of the system was used to simulate contami-nant removal for the plume at the site. The model

Water Air Soil Pollut (2013) 224:1741 Page 3 of 6, 1741

simulates rate-limited diffusion of contaminant fromlower-permeability units (in which no advection isassumed to occur) into higher-permeability zones, inwhich advective transport occurs. Details of the modelare presented in Brusseau (1991). The simulation pro-vides a reasonable representation of the measured data(see Fig. 1), especially considering the simplificationsemployed (e.g., uniform pumping rate, uniform systemthickness, uniform intra-unit permeabilities). The rela-tively good match may indicate that diffusion of con-taminant from lower-permeability units (back diffu-sion) may be contributing to mass-removal constraintsat the site. However, because of the simplified ap-proach employed, the potential impact of other factorssuch as continued mass discharge from the source zoneand well-field hydraulics cannot be ruled out, nor canthe relative significance of the various factors be eval-uated. The relative simplicity of using single- orlumped-process models for data analysis is an attrac-tive advantage of this approach. However, it is imper-ative to remain cognizant of the uncertainty and limi-tations associated with the use of such models.

The fit of the exponential source-depletion functionto the measured CMD data is presented in Fig. 1. Thecoefficients obtained from the fit of the source-depletion function are used to estimate an initial massof approximately 29,000 kg (see Fig. 1 caption forexplanation). This compares to a value of approximate-ly 20,000 kg of contaminant mass that has been recov-ered to date from the pump-and-treat operation. Thedetermination of an estimate of initial contaminant

mass allows quantitative assessment of remediationperformance and evaluation of time scales for opera-tion. The results of such assessments must be evaluatedwith due consideration of uncertainties associated withthe estimates produced with the method.

The CMDR–MR profile determined for the site ispresented in Fig. 2, along with three reference profiles.It is observed that the profile for the measured dataresides primarily above the one-to-one line. The natureof the profile is indicative that a substantial amount of

0.1

1

10

100

0 5 10 15 20

TC

E C

MD

(kg

/d)

Time (year)

Measured

Model Sim

S-D Function

Fig. 1 Contaminant massdischarge profile obtainedfrom analysis of pump-and-treat data for the selectedsite. A simulation (modelsim) produced with a dual-porosity mathematical mod-el (described in the text) isalso presented. The fit of afirst-order exponentialsource-depletion (S-D)function is also presented(CMDt/CMD0 = exp(-0.14 t);r2=0.80). Initial mass isdetermined asM0 =CMD0/k, withCMD0 = 11 kg/d

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

no itc

ude

Re

grahc si

Ds sa

Mn

oit carF

Fraction Mass Reduction

1:1

Minimal Reduction

Maximal Reduction

Site Data

Fig. 2 Plot of the relationship between the reduction in CMDand the reduction in contaminant mass. Also shown are threereference profiles

1741, Page 4 of 6 Water Air Soil Pollut (2013) 224:1741

contaminant mass remains at the site that is poorlyaccessible to groundwater flushing associated withthe pump-and-treat system (e.g., Jawitz et al. 2005;Brusseau et al. 2008, 2013; DiFilippo et al. 2010;Christ et al. 2010). The removal of this mass is thusconstrained by mass-transfer processes, which likely isat least partially responsible for the asymptotic

behavior observed for the temporal CMD profile(Fig. 1). The CMDR–MR relationship can be plottedin equivalent log form to better visualize behaviorunder asymptotic conditions (see Fig. 3).

5 Summary

The ultimate goal of remedial actions is to reduceoverall risk posed by the site, which is typically medi-ated by the groundwater contaminant plume. The stan-dard method for assessing remediation performance isbased on analysis of changes in contaminant concen-trations for groundwater samples collected from mon-itoring wells located within the treatment zone. It isadvantageous to employ time-continuous contaminantmass-discharge data obtained from analysis of pump-and-treat operation data to enhance these assessments.An example of such an application is presented inFig. 4, wherein the impacts of two major source-zoneremediation efforts (soil vapor extraction and in situchemical oxidation) are evaluated using historicalpump-and-treat data collected for a site in Tucson,AZ, USA (Brusseau et al. 2011b).

Collection and analysis of data for individual wellsfor systems with multiple extraction wells provides anopportunity to characterize spatial variability of prop-erties, conditions, and behavior in the areal plane.

Mass D

ischarg

e Red

uctio

n (%

)

Mass Reduction (%)

1:1

Minimal Reduction

Maximal Reduction

Site Data

90 99 99.9

99.9

99

90

99.99

Fig. 3 Plot of the relationship between the reduction in CMDand the reduction in contaminant mass-log version. Also shownare three reference profiles

Fig. 4 Impact of source-zone remediation on com-posite CMD measured at thegroundwater contaminantplume scale for a site inTucson, AZ, USA. Soilvapor extraction (SVE) wasstopped in year 17, and insitu chemical oxidation(ISCO) was stopped pastyear 19. The figure wasadapted from Brusseauet al. (2011b)

Water Air Soil Pollut (2013) 224:1741 Page 5 of 6, 1741

However, standard extraction wells for pump-and-treatsystems are designed (e.g., large screened intervals)such that the flow and contaminant concentration datacollected represent composite, vertically averagedvalues. It is proposed that modifying fully screenedextraction wells to collect vertically discrete flow andcontaminant concentration data would produce signif-icant benefit by providing vertical resolution of con-taminant distributions, mass discharge, and mass re-moval. A major advantage of this proposed modifica-tion is that the vertically discrete monitoring of extrac-tion wells in conjunction with the use of multiple ex-traction wells provides a three-dimensional characteri-zation of the treatment zone.

The results presented herein illustrate that compre-hensive analysis of historical pump-and-treat data is apowerful, cost-effective method for providing higher-resolution, value-added characterization of contami-nated sites. Advantages of the method include thefollowing: (a) use of data that typically exist for oper-ating sites, thus minimizing data collection costs; (b)no disruption of the operating remediation system; and(c) ability to update the analysis at any time, providinga means to periodically revise the site conceptual mod-el and optimize the operation of the remediation sys-tem. It should be noted that this method can also beused for soil vapor extraction systems (Brusseau et al.2010). It is anticipated that implementation of thisapproach will enhance decision making concerningremedy modification, long-term site management,and closure.

Acknowledgments This research was supported by the USDepartment of Defense Strategic Environmental Research andDevelopment Program (ER-1614) and the National Institute ofEnvironmental Health Sciences Superfund Research Program(ES04940). The author thanks Zhilin Guo for assistance incompiling the data and the reviewers for their constructivecomments.

References

Basu, N. B., Rao, P. S. C., Poyer, I. C., Nandy, S., Mallavarapu,M., Naidu, R., et al. (2009). Integration of traditional andinnovative characterization techniques for flux-based as-sessment of dense non-aqueous phase liquid (DNAPL)sites. Journal of Contaminant Hydrology, 105, 161–172.

Brusseau, M. L. (1991). Application of a multiprocessnonequilibrium sorption model to solute transport in a

stratified porous medium. Water Resources Research, 27,589–595.

Brusseau,M. L., Nelson, N. T., Zhang, Z., Blue, J. E., Rohrer, J., &Allen, T. (2007). Source-zone characterization of achlorinated-solvent contaminated Superfund site in Tucson,AZ. Journal of Contaminant Hydrology, 90, 21–40.

Brusseau, M. L., DiFilippo, E. L., Marble, J. C., Oostrom, M.(2008). Mass-removal and mass-flux-reduction behaviorfor idealized source zones with hydraulically poorly-acces-sible organic liquid. Chemosphere, 71, 1511–1521

Brusseau, M. L., Rohay, V., & Truex, M. J. (2010). Analysis ofsoil vapor extraction data to evaluate mass-transfer con-straints and estimate source-zone mass flux. GroundWater Monitoring and Remediation, 30, 57–64.

Brusseau, M. L., Carroll, K. C., Allen, T., Baker, J., DiGuiseppi,W., Hatton, J., et al. (2011a). The impact of in situ chemicaloxidation on contaminant mass discharge: linking source-zone and plume-scale characterizations of remediation per-formance. Environmental Science and Technology, 45,5352–5358.

Brusseau, M. L., Hatton, J., & DiGuiseppi, W. (2011b). Assessing theimpact of source-zone remediation efforts at the contaminant-plume scale: application to a chlorinated-solvent site. Journal ofContaminant Hydrology, 126, 130–139.

Brusseau, M. L., Matthieu, D. E., III, Carroll, K. C., Mainhagu,J., Morrison, C., McMillan, A., et al. (2013). Characterizinglong-term contaminant mass discharge and the relationshipbetween reductions in discharge and reductions in mass forDNAPL source areas. Journal of Contaminant Hydrology,49, 1–12.

Butcher, J. B., & Gauthier, T. D. (1994). Estimation of residual denseNAPL mass by inverse modeling. Ground Water, 32, 71–78.

Christ, J. A., Ramsburg, C. A., Pennell, K. D., & Abriola, L. M.(2010). Predicting DNAPL mass discharge from pool-dominated source zones. Journal of ContaminantHydrology, 114, 18–34.

DiFilippo, E. L., & Brusseau, M. L. (2008). Relationship betweenmass flux reduction and source-zone mass removal: analysisof field data. Journal of Contaminant Hydrology, 98, 22–35.

DiFilippo, E. L., Carroll, K. C., & Brusseau, M. L. (2010).Impact of organic-liquid distribution and flow-field hetero-geneity on reductions in mass flux. Journal of ContaminantHydrology, 115, 14–25.

Falta, R. W., Rao, P. S. C., & Basu, N. (2005). Assessing theimpacts of partial mass depletion on DNAPL source zones I.Analytical modeling of source strength functions and plumeresponse. Journal of Contaminant Hydrology, 78, 259–308.

Interstate Technology and Regulatory Council (ITRC). (2010).Use and measurement of mass flux and mass discharge.www.itrcweb.org.

Jawitz, J. W., Fure, A. D., Demy, G. G., Berglund, S., & Rao, P.S. C. (2005). Groundwater contaminant flux reductionresulting from nonaqueous phase liquid mass reduction.Water Resources Research, 41, 10408–10423.

Zhang, Z., & Brusseau, M. L. (1999). Nonideal transport of reactivesolutes in heterogeneous porous media: 5. Simulating regional-scale behavior of a trichloroethene plume during pump-and-treatremediation.Water Resources Research, 35, 2921–2935.

Zhu, J., & Sykes, J. F. (2004). Simple screening models of NAPLdissolution in the subsurface. Journal of ContaminantHydrology, 72, 245–258.

1741, Page 6 of 6 Water Air Soil Pollut (2013) 224:1741