evaluated by focusing on the main technical elements and their significance with respect to resource

funnel construction and by minimising the steel consumption for the gate construction. Granular activated

carbon (GAC) is exclusively considered as the treatment material, both in-situ and on-site. Here it is

Journal of Contaminant Hydrology 83 (2006) 171199

www.elsevier.com/locate/jconhyddepletion and potential adverse effects on ecological quality, as well as on human health. Seven impact

categories are distinguished to address a broad spectrum of possible environmental loads. A main point of

discussion is the reliability of technical assumptions and performance predictions for the future. It is

obvious that a high uncertainty exists when estimating impact specific indicator values over operation times

of decades. An uncertainty analysis is conducted to include the imprecision of the underlying emission and

consumption data and a scenario analysis is utilised to contrast various possible technological variants.

Though the results of the study are highly site-specific, a generalised relative evaluation of potential

impacts and their main sources is the principle objective rather than a discussion of the calculated absolute

impacts. A crucial finding that can be applied to any other site is the central role of steel, which particularly

derogates the valuation of FGS due to the associated emissions that are harmful to human health. In view of

that, environmental credits can be achieved by selecting a mineral-based wall instead of sheet piles for theLife cycle assessment of active and passive groundwater

remediation technologies

Peter Bayer *, Michael Finkel

Center for Applied Geoscience, University of Tuebingen, Sigwartstrasse 10, 72076 Tuebingen, Germany

Received 17 June 2005; received in revised form 28 October 2005; accepted 10 November 2005

Available online 27 December 2005

Abstract

Groundwater remediation technologies, such as pump-and-treat (PTS) and funnel-and-gate systems

(FGS), aim at reducing locally appearing contaminations. Therefore, these methodologies are basically

evaluated with respect to their capability to yield local improvements of an environmental situation,

commonly neglecting that their application is also associated with secondary impacts. Life cycle assessment

(LCA) represents a widely accepted method of assessing the environmental aspects and potential impacts

related to a product, process or service. This study presents the set-up of a LCA framework in order to

compare the secondary impacts caused by two conceptually different technologies at the site of a former

manufactured gas plant in the city of Karlsruhe, Germany. As a FGS is already operating at this site, a

hypothetical PTS of the same functionality is adopted. During the LCA, the remediation systems are0169-7722/$ - see front matter D 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.jconhyd.2005.11.005

* Correspond

E-mail address: peter.bayer@uni-tuebingen.de (P. Bayer).ing author. Tel.: +49 7071 2973178; fax: +49 7071 5059.

identified as an additional main determinant of the relative assessment of the technologies since it is

continuously consumed.

D 2005 Elsevier B.V. All rights reserved.

Keywords: Life cycle assessment; Funnel-and-gate; Pump-and-treat; Remediation; Groundwater; Uncertainty; Monte

Carlo; Weighting triangle

1. Introduction

Groundwater is one of the most valuable natural resources. Beside its central position for

terrestrial and aquatic ecosystems, groundwater is essential for drinking and industrial water

abstraction. It represents by far the largest reservoir of fresh water, with stored volumes over fifty

times greater than the amount of surface water. Growing populations, continual urbanisation and

industrialisation are increasing the risk of groundwater contamination, while simultaneously, the

living standards and demands on our natural environment are rising. Hazardous waste disposal

facilities, oil refineries and chemical plants are typical sources of waste streams, which can reach

the aquifer and produce long-term contaminations in the subsurface. Most contaminant sites in

industrialised countries are characterised by the occurrence of hazardous organic chemicals in

the underground. They can form separate, nonaqueous phases in the subsurface, which

continuously feed the passing groundwater with contaminants. Downgradient of these source

zones, plumes develop with significant concentrations of the contaminants dissolved in the

groundwater. Contaminants can also slowly diffuse into the low permeable aquifer matrix and,

because of the small mass transfer rates into groundwater, stay there as long-term secondary

sources (Grathwohl, 1998; Fetter, 1999). These aspects substantiate the importance of the

protection of groundwater from pollution and the control of already contaminated aquifer zones.

However, the considerably low mobility together with a typically high persistency of organic

contaminants is the crux for the development of suitable groundwater remediation technologies.

A still common and conventional bcleanupQ practice is pump-and-treat, meaning the pumpingof contaminated groundwater followed by on-site treatment. Pumping wells are installed in or in

the vicinity of the source or plume area in order to hydraulically capture the critical aquifer

zones. Experience shows that due to the limited mass transfer rates, aquifer restoration by

conventional pump-and-treat systems is, for the most part, not achievable within reasonable

timeframes (Eberhardt and Grathwohl, 2002; Stroo et al., 2003). Though the use of pump-and-

treat systems (PTS) is not the only technological option, it still seems to be the most favourable

because of the experience in appropriate hydraulic design, as well as its flexibility and simplicity

(US EPA, 1996; Bayer, 2004).

Meanwhile, permeable walls (PRBs) are a widely accepted alternative for long-term plume

management. Continuously or locally reactive vertical walls are installed in-situ for down-

gradient capture and treatment of the contaminated plume (US EPA, 2002). Funnel-and-gate

systems (FGS Starr and Cherry, 1994) are a variant based on the combination of impermeable

walls (funnels) and reactive zones (gates). Adjusted to the regional groundwater flow regime, the

funnels direct the contaminated water through the in-situ treatment facilities within the gates.

Compared to PTS, FGS and continuous permeable walls are denoted as bpassiveQ systems since,after proper installation of the technological devices, no active work, such as pumping, is

needed.

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199172In the presented study, the focus is set on the comparison of PTS and FGS, as typical active

and passive groundwater remediation options, in terms of their environmental impacts. Similar

to wastewater treatment or recycling techniques, the achievement of a local environmental

benefit, i.e. the control and removal of contaminants, is compromised by certain environmental

burdens caused during the construction, installation and maintenance of the technologies. These

so-called secondary impacts (Volkwein et al., 1999) can reach significant levels depending on

the type of materials and energy used, the size of the treatment system and its time of operation.

A life cycle assessment (LCA) framework is used to contrast the potential impacts in different

categories reflecting the inherent adverse effects on ecology, human health as well as the

depletion of energy resources. LCA considers the potential impacts associated with the supply

chains throughout a products (i.e. here, technologys) life cycle from raw material acquisition

through production, use and disposal (ISO, 14040, 1997).

2. Previous work

The most comprehensive approaches for using LCA to environmentally rate impacts from

remediation of contaminated sites have been presented by Beinat et al. (1997), Volkwein et al.

(1999) and Diamond et al. (1999). Sue`r et al. (2004) give a brief literature review of the few

available studies in this area. The REC method presented by Beinat et al. (1997) is a streamlined

decision support system for estimating the relevance of environmental impacts of technological

alternatives within a risk assessment framework. Volkwein et al. (1999) emphasize a difference

in their UVA approach to REC: While the REC tool can be applied to derive suitable clean-up

levels, the levels have to be assigned a priori when using the UVA method. Aside from this,

UVA is a more detailed and elementary approach, which operates on generic datasets processed

in over 50 modules (LfU, 1998). Volkwein et al. (1999) show the application of their method in

accessing the impacts caused by partial soil excavation, on-site ensuring and surface sealing.

Bender et al. (1998) also discuss the suitability of the UVA tool for analysing the impacts caused

by a number of different groundwater remediation technologies, such as long-term extraction of

groundwater, in-situ bioremediation and soil vapor extraction. Though their case study does not

exhibit a detailed site and inventory data description and no information is given on how the site

data is processed within the LCA framework, general conclusions are derived when comparing

the different remediation technologies. For example, they identified energy consumption as the

major cause of environmental impact when long-term groundwater extraction is considered. This

is supported by the findings of Vignes (2001).

The life cycle management approach of Diamond et al. (1999) is intended as a systematic

concept for deriving qualitative comparisons of selected soil and groundwater remediation

options within a decision making process. It is complemented by a LCA step for a detailed

investigation of the environmental burdens associated with the technologies. The generic

remediation options highlighted were excavation and off-site disposal, in-situ bioremediation,

soil washing, vapor extraction and no action. The utilisation of the conceptual outline presented

by Diamond et al. (1999), for a quantitative inspection of associated environmental burdens, is

shown in a case study on soil remediation of a PAH contaminated site in Ontario, Canada (Page

et al., 1999). During the examined time horizon of 75 weeks, contaminated soil and sludge was

discarded at nearby disposal facilities. Since no long term measurements such as post-site

processing and maintenance of waste landfill were investigated, the main emissions were

controlled by material transport. The creation of solid waste and land use was accounted for in

separate life cycle impact categories. Soil remediation is also subject to the LCA presented by

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199 173Blanc et al. (2004), where four technologies are compared in terms of their environmental

impacts and the related energy consumption to clean-up a sulfur contaminated site in France.

The four remedial actions work with soil excavation, followed by on-site treatment, backfill and/

or disposal.

3. LCA framework

3.1. Overview

The structural and procedural components of LCA are determined by the international

standard series ISO 1404043 (1997, 1998, 2000a,b). Mandatory initial steps include the

definition of goal and scope, functional unit as well as system boundaries. The overall goal for

this study is the comparison of the environmental performance of two different long term

groundwater remediation technologies. A premise for a valid comparison is to presume an equal

functionality of both approaches, i.e. to define the functional unit so that it provides a reference

to which the inputs and outputs can be related (Rebitzer et al., 2004). This is addressed by

analysing an existing case, the former manufactured gas plant site of the city of Karlsruhe in

southern Germany (site description given in Table 1). In 2000, a FGS (Fig. 1) was installed

(Schad et al., 2000), which will subsequently be judged against a hypothetical PTS. The latter is

Table 1

Technical specifications of Karlsruhe site and key elements of FGS and PTS subject to LCA

Site parameterformer gas plant Karlsruhe

Width of contamination 210 m

Hydraulic conductivity log mean 3.9E03 m/sHydraulic conductivity log variance 1.31

Regional gradient 0.070.135% gradient direction

(25% seasonal variation) NW/WNW

Contaminant concentrations Acenaphthene 400600 ug/l

Averaged total flow rate through FGS 10 l/s

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199174FGS

Funnel depth 17 m

Funnel length 240 m

Gates depth 17 m

Gate borings diameter 2.50 m

Gate casings diameter 1.80 m

Steel/gate 812 tGravel/gate (50 m3/gate; 1.6 t/m3) 880 tClay/gate (8 m3/gate; here 2 t/m3) 816 tGAC/gate (25 m3/gate; 0.6 t/m3) 815 tLifetime (regeneration interval) 5 years

PTS

Number of wells 10

Installation depth 17 m

Pumping rate 18 l/s

PVC conduits 100 m

HDPE container (5 cm wall) max. 20 m3 each

GAC refill interval 1 year

GAC volume per fill 2443.2 t/fillLifetime of conduits, pumps, vessels 10 years

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199 175configured assuming equivalent remediation goals, i.e. the long term hydraulic containment of

the contaminated plume emanating from the upgradient source zone. Based on the site-specific

results of the LCA, the goal is to identify crucial and environmentally sensitive technical design

elements, as well as to derive suggestions for technological enhancements for the particular site

and in general. The exact definition of functional unit used for this study is explained in detail in

the following section.

The system boundaries for the LCA are set based on the characteristics of the site, focusing

on the environmental burden related to major technological elements, which will be delineated in

Fig. 1. Scheme of funnel-and-gate system as installed at the former gas plant Karlsruhe-Ost. Vertical sheet piles act as

funnels that direct the contaminated groundwater into (in total 8) gates. In these gates, perforated steel cylinders are filled

with GAC that is regenerated every 5 years.detail after the subsequent description of the site. The presented study is not limited to the

analysis of the installed FGS and a hypothetical PTS. Several modifications of the reference

scenario are formulated during the evaluation of the results in order to investigate the role of

assumptions made for the technological elements and to develop system alternatives with

decreased environmental impact. An overview is in Table 2. The scope of these scenarios is

illustrated in Fig. 2.

After the initial phase of the LCA, which is crucial to establish the context in which the

evaluation is to be made, the ensuing steps are inventory analysis, impact assessment and

interpretation. The inventory analysis (LCI) examines and compiles all relevant energy and

material inputs and outputs of processes during the life cycle of a product or a service. The impac

Table 2

Variations between scenarios developed during the LCA

Scenario Time (years) FGS PTS

Reference scenario sheet piles GAC 72t/y

Scenarios with different wall technologies 30 diaphragm wall, slurry wall

Scenarios with different wall technologies

and variable operation time

1050 sheet piles, diaphragm wall,

slurry wall

Scenarios with diff. wall technologies,

variable operation time and variable

GAC consumption for PTS

1070 sheet piles, diaphragm wall,

slurry wall

GAC var.t

r

assessment (LCIA ) is carried out to translate the collected emissions and consumptions into

environmental and/or health effects and is commonly expressed by a number of representative

indicators. Finally, within the interpretation phase, the results of inventory analysis and impact

assessment are discussed to extract the main sources of environmental burden and to derive

recommendations for the preferable product. In the next sections, it is described in detail how each

crucial LCA element has been adopted for the comparison of both remediation technologies.

4. Functional unit

The functional unit demarcates the basis upon which services are compared within the

predefined system boundaries. In the context of groundwater and soil remediation projects the

desired service(s) may largely differ from case to case depending on the specific objectives of

the planned remedial action. At the Karlsruhe site, since a complete aquifer restoration was not

possible, neither technically nor economically, the major goal was to prevent any further

increase of risk caused by an ongoing expansion of the existing contaminant plume in the

groundwater. Both technological approaches that are in the focus of this study do achieve this

goal by hydraulic containment and a reduction of contaminant concentrations to given

standard levels by on-site or in-situ water treatment. This means that the particular

technologies, even if, in the narrower sense, they are causing different effects, do provide

Fig. 2. Key elements of FGS (left) and PTS (right) for the scenarios considered for the LCA.

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199176the same service with respect to the specific remedial goal of managing, i.e. controlling the

contaminant plume.

Consequently, the functional unit is defined here as the control of a certain contaminated

aquifer zone, which corresponds to the definition proposed by Volkwein et al. (1999). The

following arguments shall explain the reasons for the selected formulation of the functional unit,

in particular with respect to the practice in other LCA applications:

(1) The choice of this type of functional unit follows the suggestions by Shakweer and

Nathanail (2003), i.e. referring to the contaminated volume of the aquifer to be restored or

to be controlled. The performance criterion, i.e. hydraulic containment, is attained when

concentration targets are ensured downgradient of the treatment system. Models can be

applied to configure equivalent technologies, rated with respect to the remediation goals

while reflecting the present knowledge (e.g., Godin et al., 2004). In the present study we

employ an analytical model to describe the pump-and-treat system, whereas the results of

numerical modelling are utilised for the funnel-and-gate system.

(2) If remediation is defined in a broader sense, such as bimprovementQ of site situation, a moreexact functional comparison of competing actions could be to compute the trade-offs

between inputs and outputs within a risk assessment framework such as accomplished by

REC (Beinat et al., 1997). In this case, all associated environmental benefits and burdens are

balanced with each other to derive the ideal technological alternative. We do not go so far in

our study, but keep the classic definition of a service-related functional unit, which has a

priori to be set and is not a decision variable. Our objections are to the impracticably high

spatial and temporal resolution that would be needed for calculating risk-benefit trade-offs.

(3) Another limitation of the REC-approach is that its applicability and expressiveness is

restrained by the high degree of parameter and process uncertainty, since the long-term

effect and efficiency in reaching overall aquifer clean-up can hardly be estimated precisely.

There are several reasons for this: First of all, a fundamental problem is that the physical

and chemical description of the subsurface is basically inexact. Accordingly, there is

substantial uncertainty in estimating or modelling present and future status of the

remediation process. Furthermore, and even worse, at numerous sites the initial exact mass

distribution of contaminants in the subsurface is not known. Principally, descriptive and

predictive uncertainty is crucial, independent of the type of functional unit chosen.

However, the selection of a concentration threshold related functional unit is less prone to

the exact contaminant mass distribution.

(4) System design by referring to regulatory standards represents the common practice and is

compatible to economically based design considerations (e.g., Russell and Rabideau,

2000; Bayer et al., 2005a).

For a comparative evaluation of remedial options, the quality of technical design and

performance parameters fundamentally depends on the experience in technology and knowledge

of the site. If a technology is implemented at a site, its performance characteristics are known

more precisely than for another, hypothetical alternative. Also, for a realised measure, the

associated consumption in material and energy is apparent, whereas potential alternatives can be

described only theoretically. In the subsequent case study, the FGS is already installed and the

task is to set up an imaginary PTS on the same functional unit basis. As pointed out by NRC

(1997) in the example of an economic comparison of existing and hypothetical remedial options,

ones bias must be taken into careful consideration during an evaluation. Since there is extensive

experience in the application of PTS and since this technology is based on a rather simple

concept, configuring it appropriately does not represent a major difficulty. The remaining

uncertainty in design and long term performance is addressed by considering possible PTS

design alternatives in a scenario analysis.

Especially the long term trend in approaching site clean-up can hardly be predicted. This plays

a crucial role when comparing such conceptually different technologies. During the use of PTS,

(secondary) emissions are released and energy is consumed continuously, whereas for FGS, the

major impact arises during construction and installation of the system. Life cycle assessment

commonly attempts to measure the total environmental effects of a product or technology bfromcradle to graveQ (Bauman and Tillman, 2004). Obviously this can hardly be achieved when thebgraveQ, i.e. the time of termination, is not known. To overcome this, as is also common foreconomic analyses (NRC, 1997), a fixed operation period can be predefined, so that the temporal

boundary of the functional unit is explicitly set. For example, Bender et al. (1998) suggest an

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199 177operation period of 50 years for comparing long-term in-situ groundwater remediation

technologies, whereas the timeframe Page et al. (1999) assume for long term monitoring is

only half of that. It is obvious that, due to the lack of long-term experience with most remediation

technologies and the general uniqueness of site applications, these time limits are chosen quite

arbitrarily. However, timeframes shall reflect planning periods and thus also seem to be a valid

input for the LCA. In the subsequent analysis, we use time as a fixed, as well as variable,

parameter in order to reveal its influence on the comparative rating of the selected technologies.

5. Inventory set-up and impact assessment

After defining the key elements of the remediation technologies (cp. Fig. 2), an inventory was

set up in order to collect associated material and energy transfers that cross the boundary

between service system or product and environment. The configuration of key elements will be

described below when the case study is introduced. Grouping and assigning material and energy

transfers that exhibit environmental burdens and/or interventions to pre-specified categories, i.e.

characterisation and classification, is the task of life cycle impact assessment (LCIA; ISO 14042,

2000a). These categories represent threatened environmental compartments, states or relations,

such as the consumption of energy resources, global warming or human toxicity (Udo de Haes et

al., 1999; Pennington et al., 2004). For example, burning crude oil means an exploitation of

available energy resources. If this process is considered within LCA, the amount of consumed

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199178oil is assigned to the respective impact category such as bdepletion of energy resourcesQ. Burningcrude oil reduces available resources, but also releases emissions such as CO2 or SO2, which are

assigned to other categories such as bglobal warming potentialQ and bacidification potentialQ.A quantitative evaluation is only possible if the different contributions to one category,

originally expressed in different units, can be merged. This implies the necessity of a

characterisation step, which unifies all contributions and indicators. Calculation methods to

determine characterisation factors are manifold. They depend on the underlying specific models

and on the purpose they are used for. Thus there are no universally valid characterisation factors.

Especially for the human toxicity potential there is no consensus about the ideal characterisation

Table 3

Characterisation factors for selected inventory data

Impact categoryy,yy Unit of indicator Unit char.factor

Characterisation factors Ci for

inventory data

Depletion of energy resources (DER) crude oil eq.

(COE)

kg COE/kg brown coal 0.0409, hard coal 0.1836,

natural gas 0.5212

Global warming potential (GWP) CO2 eq. kg CO2/kg CH4 23, N2O 296, CF4 5700

Acidification potential (AP) SO2 eq. kg SO2/kg NOx 0.70, HCl 0.88, HF 1.60,

H2S 1.88, NH3 1.88

Terrestrial eutrophication potential (TEP) PO43 eq. kg PO4

3/kg NH3 0.35, NOx as NO2 0.13Aquatic eutrophication potential (AEP) PO4

3 eq. kg PO43/kg Nh4+ 0.327, NO3

0.095, N-comp 0.42,P-comp. 3.06, COD 0.022

Photochemical ozone creation

potential (POCP) yyC2H4 eq. kg C2H4/kg C2H6 0.189, formaldehyde 0.421,

CH4 0.007, NMVOC 0.416, VOC 0.377

Human toxicity potential (HTP) As eq. kg As/kg benzopyrene 20.9, C2H6 0.0019,

dioxine 10500, PCB 0.28, Cd 0.42,

Cr(VI) 0.28, Ni 0.06

yIndicators are calculated as Indicator Pi MiCi, where Mi (kg) is the inventory data and Ci is the characterisationfactor.yy PpIndicator of POCP is expressed nitrogen corrected as Indicator MNOx i MiCi, where MNOX compounds (UBA2000).

model, which can be based on a broad range of different health risk assessment and human

toxicology calculation methodologies. In the presented approach, the characterisation method is

adopted from UBA (2000), with the main characterisation factors listed in Table 3. A recognised

set of environmental impacts is selected (Bauman and Tillman, 2004). These are depletion of

energy resources (DER, in crude oil equivalents), global warming potential (GWP, in CO2equivalents), acidification potential (AP, in SO2 equivalents), terrestrial as well as aquatic

eutrophication potential (TEP, AEP, in PO43 equivalents), photochemical ozone creation

potential (POCP, in C2H4 equivalents) and human toxicity potential (HTP, in As equivalents).

The allocation of potential environmental impacts is not discussed with respect to spatial or

temporal variance. In this regard, we follow the underlying streamlining concept of Volkwein et

al. (1999), though, especially for conglomerate, long-term operations such as those inspected

here, a more detailed analysis seems desirable. Distinguishing between the spatially different

effects of emissions for example is discussed by Potting (2000) and Ross and Evans (2002), who

recognised the inaccuracy of geographically averaging LCIA indicators by neglecting the

regionally diverse effects of emissions. A common problem is that appropriate data sources are

fragmentary or cannot be directly transferred from the context they are gathered for.

Accordingly, no spatial discretisation of the impacts is attained here, and therefore, no answer

will be given to the question if the environmental benefit gained from local groundwater clean-

up can be balanced with the inherent global (secondary) impacts. A detailed temporal resolution

of the environmental impact was not addressed since estimations about technological evolution,

the prediction of modifications in the production of material and energy, as well as the

proceeding changes in the evaluation and allocation of environmental impacts is beyond the

scope of this paper.

In view of the variety of approaches that are competing for an apposite realisation of the

obligations according to the ISO standards, there is no unique way of conducting a LCA for

remediation technologies. Moreover, available secondary data sources and process specific

libraries, which have been preferred for the presented study, can exhibit significant

discrepancies depending on sub-system boundaries, desired temporal and spatial resolution,

specific calculation models used and actuality. This inherently means that the expressiveness

of the results can only be judged by reflecting the validity of the input data. The LCA

framework developed in this study is oriented at and applied to an example case in Germany,

so also the focus was set on exploiting (secondary) data bases with close relation (partly only

available in German), and adopting the standards recommended for LCA in other disciplines

by the German environmental agency (e.g., UBA, 1998, 2000). Due to the fact that not solely

case-specific but partly also pre-existing approaches are used for the description of

interconnected sub-processes involved (e.g., production of electrical energy, oil, steel), the

accuracy of the calculations for the specific case study is limited. However, this presented

streamlined concept is not aiming at an exact quantification of the potential environmental

burden of the technologies considered. Instead it shall constitute similar boundary conditions

for the description of the technologies and this way shall enable a valid comparative

assessment. In order to incorporate the expected impreciseness of inventory data into the

presented LCA, ranges instead of deterministic values are considered, and these are processed

within a subsequent uncertainty analysis.

An optional step in LCA is a cumulative investigation of all categories together. It can be

used for a more condensed presentation of the results. In general, aggregation of impacts

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199 179should be done carefully in order to avoid a loss in transparency concerning the role of

individual impacts. A critical effect is that it can lead to severe compensation between impact

results (a good impact results can compensate a very bad one), and so it is recommended to

discuss the results for individual impacts before the aggregation. Subsequently, the two

remediation technologies are contrasted by means of a scenario analysis and it seemed

desirable to partly aggregate categories for the presentation of the results. We group the

categories into three types: those denoting adverse effects on energy resources (DER),

ecosystem quality EQ (GWP, AP, TEP, AEP) and human health HH (POCP, HTP). This step

enables a basic examination of the role of crucial parameters and visualisation by mixing

triangles presented by Hofstetter et al. (2000), which also suggested the distinction of these

three safeguard subjects according to the Eco Indicator 99 methodology. The aggregation into

EQ and HH classes is carried out by averaging the category-specific indicator values. First, in

order to achieve a comparability between the different kinds of impacts, the indicator values

have been normalised with respect to (annual) inhabitant equivalents (Table 4). By

normalisation, the indicators represent the category specific multiple of the prevailing amount

Table 4

Impact assessment for key elements of FGS and PTS

Inhabitant value/year Units DER

(kg COE)

GWP

(kg CO2)

POCP

(kg C2H4)

AP

(kg SO2)

TEP

(kg PO43)

AEP

(kg PO43)

HTP

(kg As)

2447 13167 56 5.7 7.8 14 0.006

GAC ton 904.8 10975.1 1.21 5.83 0.52 1.1E03 4.2E05GAC rec. ton 328.3 1166.6 0.54 1.76 0.30 0 3.1E05GAC transport ton 34.8 110.8 0.24 0.65 0.13 0 1.4E05Facilities 97.9 305.5 2.52 0.25 0.55 1.4E07 4.2E05

FGS WALL

Sheet pile m2 44.5 363.3 1.02 0.20 0.24 2.8E03 5.7E03Sheet pile + rec. m2 19.5 164.9 0.59 0.10 0.14 7.4E04 1.4E03Facilities 562.8 1799.5 14.4 1.43 3.07 0 4E05Diaphragm wall m2 7.25 28.3 0.25 2.7E02 6.0E02 1.1E10 1.7E06Facilities 765.3 2336.7 26.5 2.13 4.80 0 2.9E04Slurry wall m2 18.6 139.9 0.51 6.8E02 0.12 1.4E10 7.4E06Facilities 1757.7 5320.4 68.7 5.70 13.0 0 5.0E04

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199180FGS GATE

Excavation m3 23.0 156.8 0.42 0.33 0.05 0 0

Facilities 470.3 1400.6 20.6 0.84 1.92 0 0

Rebuild m3 4.14 12.7 0.14 1.4E02 3.2E02 5.9E09 1.2E06Facilities 89.5 281.6 2.18 0.22 0.49 1.3E07 3.7E05Steel ton 284.8 2596.5 5.13 1.22 1.23 2.1E02 4.2E02Steel + rec. ton 124.6 1178.3 2.99 0.62 0.70 5.6E03 1.1E02Gravel ton 6.56 20.3 0.17 1.4E02 3.0E02 0 3.1E06Clay ton 27.9 77.9 0.39 0.06 9.4E02 0 1.1E05

PTS

Installation well (17 m) 146.3 351.1 3.71 0.37 0.58 1.3E03 2.2E05Facility 12.56 42.46 0.23 3.2E02 3.8E02 0 4.6E06Pumping 10E+6 L 6.68 96.7 7.3E02 1.7E02 2.4E02 1.5E04 2.4E06Pump 2.35 23.1 4.9E02 1.2E02 1.1E02 2.1E04 1.2E04Conduits 100 m 2.52 7.95 4.0E02 6.4E03 6.9E03 2.2E04 1.6E08GAC container vessel (20 m3) 1340.0 2243.2 20.6 2.52 2.19 2.2E02 4.8E06

Average inhabitant emissions and consumptions for Germany are taken from IFU (2001).

of emissions or consumption rates. Hence, based on the same units, they reflect the relative

additional burden caused by the technologies under consideration. However, they do not

reflect the level of seriousness adherent in the individual categories.

The LCA framework, including material flow and interrelations of individual process steps

describing the life cycle of the remediation technologies, has been developed within the

modelling environment UMBERTO, a commercial LCA software (IFU, 2001). It represents a

widely used modelling framework, which enables material flow analyses based on comprehen-

sive state-of-the-art libraries (material groups, process data) that can be expanded depending on

the specific problem requirements. After introducing the site in the following section,

assumptions on processes involved and on descriptive parameters for the case study are

exposed. A detailed list of the inventory data is given in ZAG (2003).

6. Interpretation

An elementary and not an exclusively final step of LCA is the interpretation phase. As

demonstrated here, it shall accompany the course of LCA. Its purpose is to analyse results, reach

conclusions, explain limitations and provide recommendations based on the results of the

preceding steps (ISO 2000). In particular, interpretation shall identify and discuss environmen-

tally most relevant elements of the remediation technologies, while reflecting uncertainty in

inventory data and technical assumptions. Ideally, it is possible to derive suggestions for

technological improvement with respect to the different LCA impact categories.

7. Case studyformer gas plant site, KarlsruheOst

In the year 2000, a FGS was installed downgradient of a PAH dominated circa 200 m thick

plume emanating from the former manufactured gas plant site of the city of Karlsruhe in the

Rhine valley (Fig. 1). The site covers an area of about 100,000 m2 (Schad et al., 2000) with

several supposed source zones of tar oil. The slightly confined aquifer is 12-m thick, underlain

by a clay layer at a depth of 16 m below the surface. Due to the high conductivities of the

alluvial gravel dominated deposits of the Rhine, the contaminated groundwater flowrate

emanating from the site is estimated to reach 12 l/s under natural conditions. The highest

concentrations of the contaminants in the plume are measured for Acenaphthene (600 Ag/l), withbenzene and vinyl chloride as secondary pollutants. The latter is suspected to occur in the aquifer

upgradient of the former gas plant and thus originates from another source.

The FGS consists of 240 m funnel and eight gates, which entirely capture the plume. The

funnels were constructed of sheet piles, which are installed at a depth of 17 m. For the gates, the

excavation of borehole casings with diameters of 2.5 m were needed throughout the entire

aquifer depth. Then cylindrical gate segments, each being 3 m in length and 1.80 m in diameter,

were installed. These are pre-fabricated out of steel that is perforated at the in- and outlet and is

connected to the aquifer by gravel fillings. At each gate, clay sealings connect the sheet pile and

steel cylinder segments (Schad et al., 2000; ZAG, 2003). After the construction of the permanent

facilities, a total mass of 120 t (200 m3) granular activated carbon (GAC) has been filled into the

steel cylinders. With a reduced flowrate of 10 l/s, the GAC lifetime is estimated at 5 years. This

period is also set as a reference value for the FGSs GAC exchange and regeneration cycles for

the LCA (cp. Tables 1 and 2).

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199 181The dimensioning of an equivalent hypothetic PTS depends on the groundwater flowrate, the

area to be captured, the type and concentration of contaminants, as well as regulatory restrictions.

Here we assume that the required (total) pumping rate can be determined by referring to the

flowrate of contaminated water. We use a rough estimation according to Javandel and Tsang

(1986), which present type curves for steady-state well capture zones in confined, homogeneous

aquifers. Following their approach, the pumping rate to capture a certain area is calculated as a

multiple of the uniform flowrate, which is 12 l/s at the Karlsruhe site and serves as a lower limit. In

our study, we set 18 l/s as a reference. This value is selected with the assumption that downgradient

pumping has to be moderately higher, but not more than double, the uniform rate to assure

complete hydraulic capture (Bayer et al., 2003). The ideal number of pumping wells to be installed

is uncertain, but due to the high extraction rate, it is assumed that in total, ten fully penetrating wells

have to be installed. According to the in-situ treatment in the gates of the FGS, GAC is also chosen

as a reactive media for PTS on-site. The cleaned-up water is disposed of in local sewers or directed

to the Rhine river. Though there is a chance that local regulations impose further processing of even

decontaminated water in local wastewater facilities, this is beyond the scope of this paper and is not

considered within the boundaries of the LCA.

8. Classification of technological elements considered within LCA

8.1. LCA technical boundaries

The focus of this comparative study is set on the facilities installed and maintained for each

particular remediation technology. The FGS and PTS facility components chosen for the LCA

are expected to represent the core elements, in terms of their environmental relevance. It is

assumed that additional services not mentioned in Table 2, such as long term monitoring and

sampling, represent only a lower ontribution to the LCA results (cp. Page et al., 1999) and

therefore can be neglected for both technologies. It is obvious that several further processes and

elements could be considered, which together may have a noticeable influence on the final LCA

based comparison. However, most of these are highly site-specific (personal activities, local

power control facilities, etc.), very variable and/or can only be assessed through a

disproportional effort to gather reliable data. Therefore, and in order to draw general conclusions

from the presented assessment, no further work has been done in this direction. More significant

impacts can be expected from required technical modifications because of unsatisfactory system

performance. For example, the FGS in Karlsruhe was slightly revised during the first years since

its implementation and capping elements were installed at the top of the gates to improve its

hydraulic performance. However, considering these modifications seemed unreasonable, and it

is not possible to develop a similar scenario for the PTS alternative. Despite this, it should be

emphasised that the assumption of no modifications and no technical failure can be principally

realistic, but realistically is a very optimistic assumption.

9. Upstream processes

The process flow diagrams of the main technological elements are depicted in Fig. 3. There

are a number of basic separable processes that deliver inputs at several positions and are

distinguished as upstream processes within the modelling framework. These are in particular the

supply with diesel and electrical energy. For these, the available UMBERTO libraries are used,

which are based on the inventory data provided by Frischknecht et al. (1996). Electrical current

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199182is consumed by processes such as steel construction and slurry mixing. The percentage

contribution among different generation sources is set according to IFU (2001), with 32.5%

nuclear power, 27.0% hard coal, 27.3% brown coal, 7.6% natural, 3.9% water power gas, 0.5%

fuel oil, 0.3% wind and 0.9% waste. The quality of the inventory data on diesel (fossil fue

refinement, production) and electrical current can be judged as sufficient for the purpose of this

study since, despite its age, the database is widely used and 27 comprehensive.

One further common process is the usage of machinery, such as for site preparation, drilling

and (de-)installation of technical devices and transport activities. Please note that the fabrication

of this machinery was not considered, only the energy consumptions and emissions during

operation. The diesel consumption and combustion emissions are calculated the same for al

devices according to BUWAL et al. (1994) and Borken et al. (1999). Utilising the empirica

approximations provided by BUWAL for machines of capacities above 150 kW, we estimated a

Fig. 3. Flow sheet of the scope of the LCA for the key elements of the remediation technologies. Circular arrows indicate

temporal replacement of elements during operation.l

lP. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199 183l

demand for diesel of 242 g/kW h, and emissions of 14.3 g/kW h NOx, 0.96 g/kW h VOC, 2.4 g/

kW h CO and 1.0 g/kW h dust. These values are utilised for machines used for the installation of

walls (silent piler, clamshell) and reactor excavation. The emissions from the usage of diesel

driven devices with capacities between 25 kW and 150 kW are also approximated by the

empirical functions recommended by BUWAL et al. (1994).

Another ubiquitous process is the transport of machines to the site and back, the transport of

mineral substances (e.g., gravel, sand, clay excavated soil), as well as of technology-specific

equipment and material (e.g., sheet piles). For the allocation of the diesel consumption and

emissions caused from transport activities, the internal UMBERTO module was used, which is

based on the TREMOD (traffic emission estimation model) by Borken et al. (1999).

The underlying data quality on logistic processes is not very good and, especially due to its

deficient up-to-dateness, the quantitative description has to be used with caution. A preceding

sensitivity analysis (ZAG, 2003), however, indicated only a minor relevance of transport activities

on the entire environmental burden for the selected remediation technologies. In particular, a

scenario analysis with reasonable min/max assumptions for transport distances and inventory data

revealed that variations lead to changes of the impact category specific indicators below 10%.

10. Vertical barriers

The key elements of the LCA for FGS are listed in Table 1: funnel, gate and GAC. For the

funnel, sheet piles are implemented over a total area of 240 m17 m=4080 m2. The inventorywas set up utilising secondary databases and according to the information of the manufacturers

(ZAG, 2003). For the funnel construction, a number of different wall types are available, such as

slurry walls, thin diaphragm walls or sheet piles. The choice usually depends on several factors,

including depth, length, stability, longevity and economic considerations. A first inspection of

the category related to emissions and consumptions, as listed in Table 2, reveals that significant

benefits per m2 wall can be achieved by installing an alternative barrier such as a (thin)

diaphragm wall or a cement bentonite slurry wall. This observation motivated a comparison of

the existing sheet pile type to these alternatives, which are potentially associated with a lower

environmental burden. Please note that a number of other types of walls exist, such as sheet piles

of different material (aluminium, precast concrete), soil-bentonit slurry walls, bore pile or

composite cutoff walls (Meggyes and Simon, 2000; Carey et al., 2002).

The funnel at the Karlsruhe site was constructed out of pre-fabricated steel strips, which were

successively pressed into the soil, keyed in the underlying aquitard and joined to form a

continuous subsurface barrier. Diesel consuming press-in machines (here: silent piler, provided

by GIKEN) were used to push the wall segments into the ground. The silent piler specific

nominal capacity is set to 650 kW at 25 m2/h net progress when installing the piles. As a

preparatory measure, a hydraulic driller (200 kW at 30 m2/h) was utilised to loosen the ground

and enhance the subsequent pile countersink. The transport distances of the sheet piles were

assumed to be 300 km at 140 kg/m2.

For manufacturing the steel strips, the process chain of steel production, inclusive raw

material mining, iron casting and forming is included. The individual processes were described

using the UMBERTO libraries. According to common practice in Germany, the steel is assumed

to be produced in parts in a blast furnace (75%) and as electric steel (25%), the latter being

environmentally preferable due to the use of discarded metal. When steel recycling is suggested

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199184after completion of the remediation, the steel is assumed to be processed in an electric arc

furnace at a volume loss of 10%. The emissions and raw material consumption for providing the

steel strips are calculated based on the data sets provided by Corradini and Kohler (1999). The

data quality is estimated as satisfactory for the purpose of this study.

The production of the two other wall types is significantly different, since they are

constructed entirely in-situ. For example, thin diaphragm walls are emplaced by using high-

pressure jetting of the slurry into the ground, as a drill stem is raised up through the ground. A

continuous barrier is produced by constructing a row of overlapping panels (US EPA, 1998;

Pearlman, 1999). As a result, slurry and energy (diesel, electric power) is consumed, and

respective site facilities (drilling devices, slurry mixing plant) have to be maintained. Transport

distances for the technological devices and machinery are assumed to be 150 km. Mineral

compounds are estimated to be transported over a distance of 20 km.

Similar to a thin diaphragm wall, a cement bentonite slurry wall is fabricated in the subsurface

by injecting bentonite-based solidifying fluids. However, for this wall type, a trench has to be

excavated in the target zone before the backfill is injected. Accordingly, a trench-cutting machine

(clamshell) has to be supplied. The excavated native soil is not assumed to be contaminated in

this study and is disposed of within the vicinity at a distance of 15 km. The transport distance for

the milling cutter machinery is held fixed at 150 km. Again, transport activities, energy and wall

material demand are considered for the allocation of the category specific inputs.

Bentonite is a clay mineral which is commonly obtained from natural deposits. Here 65% of the

bentonite is estimated to be imported from other European countries (ZAG, 2002). For the slurry

wall, the wall thickness is set to 80 cm, with a mean content of 14% (by weight) bentonite, 16%

blast furnace cement and 70% water (LfU, 1995). According to LfU (1995), 45% loss of the wall

cubature has to be expected due to solid displacement in the subsurface controlled by in-situ

sedimentation, penetration and filtration processes. The diaphragm wall is constructed out of 8%

bentonite, 9% blast furnace cement, 44% water and additional native fine sand fillers, with a

demand of 0.14 t per m2 wall. The cement is expected to be a mixture of 70:30 of metallurgical/

Portland cement, whereas this high portion of metallurgical cement shall guarantee the desired

sulfate resistance of the wall. Further information on the inventory set-up for the material used is

given in Frischknecht et al. (1996), Kohler (1999) and ZAG (2003), which evaluate the data quality

as satisfactory.

Within this study, the selected wall technologies are expected to be equal in their

performance, i.e. all exhibit the desired low permeability without failure or leakage over the

whole planning horizon. It is obvious that this is a very idealised assumption because each wall

type involves individual strengths and weaknesses. Sheet piles, for example, are especially

favourable in view of their long-term stability, chemical resistance and the related quasi-

impermeability, though leakage can occur when the separate steel strips are not appropriately

connected (US EPA, 1998; Carey et al., 2002). The experience of the Karlsruhe site indicates

that such constructional deficiencies are not present and thus will not be considered.

Accordingly, potential defects of the alternative hypothetical wall types are neglected. However,

a matter of discussion might be the longevity of the bentonite-based barriers. Though not

discussed further within the LCA framework, shrinkage cracks might occur locally, increasing

the barriers permeability. Furthermore, contaminants dissolved in the groundwater can cause a

gradual degradation of the wall (Pearlman, 1999). In contrast to sheet piles, bentonite-based

walls cannot be recycled and it is assumed that no rebuilding or restoration activities are planned

after the remediation is finished.

The impacts aggregated during the LCIA are expressed as a linear function of the wall area2

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199 185(indicator units per m ), whereas machinery transport is approximated by a fixed value for each

category. Indicator values are listed in Table 4. The most notable difference between sheet piles

and the other two variants is the huge demand for steel, resulting in relatively high-energy

consumption and a distinctively stronger rating of its HTP (Table 4). It is questionable if sheet

pile walls can be pulled out of the ground after the long planning periods of groundwater

remediation, especially if corrosion processes merge the steel with the surrounding soil matrix. If

excavating the wall is a practicable option, then the raw material gained by recycling yields

savings of 50% for the different categories (Table 4). The impact on aquatic eutrophication and

human toxicity potential even decreases by 75%. It is apparent that the unit impacts for the walls

are decided by the material they are constructed of, though they are also affected by the work

involved in (de-)installation.

11. FGS Gate

The main process steps involving gate construction, maintenance and rebuilding are:

Soil and sediment excavation: eight gates are installed, each at a depth of 17 m. Per gate, a

volume of 84 m3 soil, sand and gravel was excavated. Transport of the excavated material for

disposal is assumed to be 15 km, as described in the previous section. The nominal capacity

of the dredger is set to 70 kW at 5 m3/h net progress.

Cylindrical steel casings (Schad et al., 2000) are pre-fabricated, perforated by two thirds in the

inflow and outflow face at the tube sides. As an approximation of steel production and

processing, the same datasets are used and equivalent assumptions aremade as in the case of the

steel strips used for the sheet pile installation mentioned earlier. Transport distances from

manufacturing to installation sites are set to 200 km. After the tubes are carried to the site, a

crane lifts the steel casing segments into the gates. Approximately 12 t steel was spent for each

gate yielding a total of 96 t steel for the entire FGS. For the crane, a transport distance of 100 km

is considered.

In each gate, 8 m3 clay was used to seal the opening between funnel and gate casing.

Furthermore, 50 t gravel per gate was needed to fill the remaining voids and focus the

groundwater flow to the steel tubes, which were subsequently filled with granular activated

carbon (GAC). The treatment material will be the subject of an extra chapter.

During the FGS in operation, no further activities (except of GAC exchange) are considered for

the gate. This is a very conservative assumption, since monitoringmay cause low but long-term

emissions (well installation, periodical pumping, sampling, transportation, lab tests, etc.).

However, especially due to the relatively small expected overall contribution and the very site-

specific character of monitoring activities, they are set beyond the LCA boundaries.

It is anticipated that the gate construction will be excavated by cranes after a pre-specified

planning period. The steel will be recycled and the remaining hole backfilled at double net

progress compared to excavation. Steel recycling shall follow the same procedure as

described above for recycling the steel of sheet piles. Backfill is supplied from nearby gravel

pits at a distance of 20 km.

12. Pump-and-treat system

The core elements of the hypothetical PTS are wells with pumping devices, conduits and an

on-site treatment facility. The ten wells assumed for this case study fully penetrate the aquifer to

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199186a depth of 17 m. All wells (F 300 mm) are drilled in advance of the remedial operation. Diesel isconsumed for transport (50 km) and application of the drilling apparatus. For the latter, a

respective nominal capacity of 80 kW at 2.5 m/h net progress is assumed and the associated

emissions are approximated according to BUWAL et al. (1994). Plunged well casings (F 100mm) stabilise the borehole while allowing groundwater to enter the well. PVC is chosen as the

material for the pipes (50 kg each) used for the casings. The emissions are calculated according

to the comprehensive process chains and data supplied by Boustead (1999). The casings (density

1.5 g/cm3) are surrounded by a filter, for which gravel is used that is transported over a distance

of 20 km from nearby gravel pits.

The emissions caused by manufacturing the pumps have not been analysed in detail due to a

presumably low contribution to the LCA categories. Instead, the emissions are approximated by

the mass of steel spent on their fabrication, which is very conservatively set to 50 kg per pump.

After usage, the steel is expected to be recycled. The electricity needed for operating the pumps

is approximated using the method of Bayer et al. (2005a), assuming a degree of efficiency of

60% for the pumps (inclusive motor), and an average head loss of 20 m. We calculated a demand

of 0.15 kW h/m3 groundwater processed.

The total length of the conduits connecting the wells with the on-site treatment facility is chosen

as 100 m. For the fabrication of the conduits (F 100 mm), a total mass of 300 kg PVC is estimated.The conduits direct the extracted groundwater into on-site vessels for which only the material

(HDPE) consumed is analysed. For the containers filled with GAC, a maximum volume of 20 m3

each and a wall thickness of 5 cm is assumed, due to road transport restrictions. A fixed value of

100 km is selected as the transport distance between the site and manufacturing location of

conduits and vessels.

The longevity of conduits, vessels and pumps is assumed to be limited to 10 years, so that

they have to be replaced every decade. Also, the productivity of pumping wells has to be

maintained, which commonly affords regenerative actions and material replacement. As a rough

approximation, to include this aspect in the LCA framework, 10% of the energy and materials

computed for the initial well construction are added in at 5-year intervals.

13. Granular activated carbon

The use of granular activated carbon (GAC) is very common for groundwater clean-up,

though only a few recent applications employ it in-situ (US EPA, 1998; Kraft and Grathwohl,

2003; Susaeta et al., 2005). It is based on a straightforward principle: the commonly

hydrophobic organic contaminants in groundwater sorb on the surface of this carbon-rich,

highly porous material when water flows through the respective GAC filters. After a certain

operation time, when the sorption capacity is exploited, the GAC filling has to be exchanged.

For long-term remediation, where treatment filters are continuously fed with contaminated

groundwater, this means a periodical exchange of used with non-used material. Ideally, used

GAC can be recycled by thermal treatment and then re-filled. There are several sources that

serve as raw material for the production of GAC, such as coal, wood or any organic waste.

Accordingly, the fabrication steps differ, for example how the raw material is produced, then

destructed and agglomerated, also affecting the final sorption capacity during water treatment.

The assumptions made for the GAC in this work are adopted from a previous study (Bayer et al.,

2005b) discussing the production of GAC out of hard coal. After pulverisation of the raw

material, a binder is added, followed by agglomeration into briquettes, drying and thermal

activation. Thereafter, the product is sieved, packaged and transported. In total, producing 1 ton

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199 187GAC affords 3 tons of hard coal, consumes 1600 kW h power and 12 tons of water vapour that is

heated by burning 330 m3 natural gas. Regenerating instead of fabricating virgin GAC is

commonly associated with less environmental impacts. Here, used GAC is heated and then

treated at high temperatures with vapour to gasify and degrade the contaminants. This

reactivation procedure usually causes a loss in material weight of about 10%.

The data quality and deepness in assessing GAC life cycle processes adopted from Bayer et

al. (2005b) is judged as fair, especially due to the hypothetical assumptions made in regard to

type and processes involved in the supply of the raw material. Since there is no unique raw

material for the production of GAC, it can be expected that, for example, in the case of using

regenerative educts instead of hard coal, particularly the influence of GAC consumption on the

depletion of energy recourses (DER) will be lower. However, the number of regeneration

intervals, which rise in the course of the long-term remediation, reduce the relative amount of

virgin GAC. Increasingly, regenerated GAC is applied, which is assumed to be produced

through the same process, independently from the original raw material.

14. Comparison of scenarios

14.1. Scenarios set-up

The set-up of different scenarios is necessary to compare the technologies depending on

variable and uncertain parameters that control system design and performance. Uncertainties

arise both with respect to the accuracy of the basic inventory data as well as to the description of

performance and duration of the remediation technologies. Additionally, for the hypothetical

PTS, the required technical adaptation to the site can not exactly be specified. To address this,

several scenarios are designed for delineating the space of possible technological variants. The

focus is set on discussing the most environmentally relevant elements and factors in order to

elaborate improvement strategies for the technical design. A main factor for the LCA outcome is

obviously the planning period, which is therefore varied among the scenarios in order to

examine its influence. Further scenarios are defined after a preceding examination of the relative

impact of key technical elements, the most significant of which turned out to be GAC

consumption and the amount of steel used. Particularly their role will be quantitatively inspected.

Aside from this bscenario uncertaintyQ (Huijbregts, 1998), there is a considerable parameteruncertainty in the underlying inventory data, i.e. the estimated emissions and consumptions,

reflecting quality and subjective interpretation of the data sources chosen. Simultaneously, the

description of the technical elements, their adoption, performance and related energy consumption

is subject to a certain degree of uncertainty, especially because a static representation of the

prevailing conditions at the Karlsruhe site is used here. This assumption is contrary to the fact that

crucial natural processes are commonly highly dynamic (e.g., groundwater flow and long term

contaminant movement in the aquifer). A more dynamic modelling and assessment method,

however, is beyond of the scope of this paper. The presented evaluation is solely based on

temporally averaging functions and therefore uses no time-dependent estimations of environ-

mental impacts. This is for the purpose of simplification and to avoid the high level of effort that

would be required to appropriately delineate the imperfect knowledge of future processes, such as

through the use of numerical groundwater flow and transport models.

The uncertainty analysis has been carried out using the Crystal Ball software (Decisioneering,

Inc.). Following the approach of Canter et al. (2002), a uniform Beta-distribution (shape

parameters a=b =2) was assumed to characterise the (potential) variability of parameters (Fig.

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 1711991884). The mean of the Beta-distribution is set to the expected value of each parameter; the range is

scaled by the degree of uncertainty of the respective parameter (F25% or F50%). The

uncertainty associated with the underlying inventory data is assumed to be F50%, reflecting amoderate impreciseness. This average value is anticipated, since exact uncertainty estimates or

Fig. 4. Scheme of beta distribution function with shape factors a =2 and b =2, scaled within uncertainty interval aroundexpected value E(X).

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199 189discussions of expected data variability are scarce. It is set slightly higher than denoted by

Volkwein et al. (1999) in order to include estimation errors when transferring consumption and

emission values to the Karlsruhe site. Inaccuracy from the imprecise determination of material

and energy consumption is assumed to be F25% for funnel and gate construction. All furtherkey elements categorised above are subject to a higher uncertainty (F50%), because they evolve(partly) in the future (e.g. ongoing supply of GAC) or are hypothetical estimates (PTS elements).

The results of the LCA carried out for the various scenarios are discussed in comparison to a

reference scenario, which shall serve as a basis for our analysis (Table 2) . Starting from this

scenario, further scenarios are developed to scrutinise the influence of these tentative

assumptions and to highlight the relevance of particular technical elements (e.g., type of

vertical wall) as well as services (e.g., sheet pile recycling).

15. Reference scenario

The reference scenario compares a FGS with sheet piles, as installed at the site, to a

hypothetical PTS, configured as discussed above with ten wells pumping at a total rate of 18 l/s.Fig. 5. Reference scenario with category-specific indicators: depletion of energy resources (DER), global warming

potential (GWP), acidification potential (AP), terrestrial as well as aquatic eutrophication potential (TEP, AEP),

photochemical ozone creation potential (POCP) and human toxicity potential (HTP). Indicators have been normalized

against inhabitant emissions or consumptions and are expressed as inhabitant equivalents (Ieq.).

The GAC consumption is assumed to increase linearly with the volume of water treated and

therefore amounts to 40 tons per year for the FGS and 72 tons for the PTS. The operation time,

i.e. planning period, is set at 30 years. To account for the parameter uncertainty, a Monte Carlo

analysis was carried out based on 10,000 samples. Fig. 5 depicts the results computed for the

individual impact categories. All potential impacts are expressed as inhabitant equivalents (Ieq.),

i.e. impacts are normalised to the category-specific expected annual emissions per inhabitant in

Germany, which are extrapolated over the operation period of 30 years (Table 4, UBA, 2000).

The Monte Carlo analysis delivers not only mean values of these indicators, but also their

probability distributions according to the variability assigned to inventory and technical

parameters. From the probability distributions, the 5% and 95% percentiles were selected to

delineate the expected ranges. These are shown as cumulative error bars for the individual

categories (Figs. 5 and 6).

For this scenario and the FGS, the impacts caused by fabricating and installing the funnel

(sheet piles) dominate in nearly all categories. The potential consumption of energy resources

and the emissions causing global warming are especially influenced by GAC due to the usage of

Fig. 6. Impacts calculated for the reference scenario assuming two alternative wall types for the FGS, a thin diaphragm

wall and a cementbentonite slurry wall.

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199190coal and its high temperature activation. The most significant impact of the funnel installation

can be observed for the human toxicity potential (HTP). The inhabitant equivalents indicate

relative emissions which are orders of magnitude higher than those calculated for the other

categories. This is particularly effected by the high emission rate of heavy metals and further

carcinogen by-products during the production of steel. Consequently, compared to the PTS, the

usage of steel turns out to be a major disadvantage for the FGS. The DER and GWP indicators

determined for PTS and FGS are nearly equal for both technologies, since the impacts from

higher consumption of GAC assumed for PTS outweigh those calculated for funnel installation.

16. Alternative wall technologies

Within this study, for the Karlsruhe site, two alternative vertical wall technologies are

suggested. This is either a slurry or a (thin) diaphragm wall. The question is whether the impacts

associated with funnel construction can be lowered if a different wall type is used. The results

are shown in Fig. 6. Compared with the huge environmental burden from usage of steel for a

sheet pile (Fig. 5), the predominantly mineral materials required for the construction of the two

other wall types yield lower impacts in all categories. The indicators calculated for the

Fig. 7. Weighting triangles balancing potential depletion of resources (DER), the expected encroachment of ecological

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199 191diaphragm and slurry wall are in the same range with slight benefits for the former, which can be

primarily attributed to lower raw material consumption. The most important effect, when

selecting a mineral-based wall type, is the reduction of HTP to less than one tenth of that

calculated for the sheet pile. The steel required for cylindrical casings installed in the eight gates

becomes a dominating factor leading to an indicator value of about 6 inhabitant equivalents.

Please note that it is assumed that the steel is recycled after the operation is ceased.

Weighting triangles (Figs. 7 and 8), oriented on the work of Hofstetter et al. (2000), represent

a condensed view of the results from LCIA. Here, only the potential depletion of resources

(DER), the expected encroachment of ecological quality (EQ) and the hazardous effects on

human health (HH) are distinguished by averaging the corresponding individual categories. Each

point within the triangle denotes a particular relative weighting of these three aspects. The

maximum weight of 100% is signified by one of the corners and it gradually decreases to 0% for

the specific aspect towards the opposite side of the triangle. Depending on the relative weighting

of the three aspects, one technology is preferable due to a lower (normalised and aggregated)

quality (EQ) and the hazardous effects on human health (HH). Figs. 6ac depict the decreasing impacts by changing the

wall type for the FGS from sheet piles (with recycling after 30 years) to slurry walls and thin diaphragm walls. Between

the preference areas for PTS and FGS, a white area of indifference delineates where no clear benefits for either

technology can be ascertained.overall impact, yielding areas of pros for PTS or FGS in the weighting triangles. These

preference areas were derived from the Monte Carlo analyses in order to incorporate the

prevailing data uncertainty. Based on these analyses, only relative weightings, which favour one

remediation technology by more than 75% of the samples, are assigned to this alternative. As a

Fig. 8. Weighting triangles for comparison between PTS and FGS at different operation times with technology-specific

preference areas. The wall type for the funnel of the FGS considered is the thin diaphragm wall.

result, an area of indifference with no clear advantage for one technology exists. This presented

concept of incorporating uncertainty by Monte Carlo Analyses into weighting triangles is, to our

knowledge, new and could be developed further by 3D diagrams or by using fuzzy membership

functions to represent the results.

When using weighting triangles, it is possible that one alternative is preferable with respect to

all categories, a trivial case which renders the overall triangle as a preference area. Such a trivial

case, for example, is the reference scenario where a sheet pile is installed and no recycling of the

steel is assumed. If the sheet pile is de-installed and recycled after the operation period of 30

years, the FGS is preferable only if the depletion of energy resources is set as the criterion. The

corresponding weighting triangle is depicted in Fig. 7a. Both PTS and FGS are indifferent with

respect to the associated ecological impacts, but the significant discrepancy in adverse effects on

human health favors PTS even if only a small weight is assigned to this aspect. As is shown in

Fig. 7b and c, the preference areas for FGS extend if a slurry wall is installed instead of a sheet

pile. The diaphragm wall reveals an even broader preference area for the FGS than for the PTS,

with benefits due especially to a minor ecological burden.

17. Influence of operation time and prediction of long-term performance

One of the most important factors for the outcome of the LCA is the planning period over

which emissions occur and over which resources are consumed. There is a crucial difference in

the effect of time on the evaluation of active and passive remediation technologies. Active

technologies such as PTS cause relatively small initial impacts (for construction and

installation), but entail continuous impacts over the entire planning period, while passive

methods, such as FGS, are characterised by a relatively high ratio of initial impacts. This is the

reason why FGS are commonly ecologically as well as economically inferior to PTS when only

considering the initial operation phase. This imbalance is expected to disappear in the long-term

with benefits evolving for the FGS. The role of time for the LCA of the remediation technologies

in question is further scrutinised by the weighting triangles shown in Fig. 8ac. The wall type

with the best LCA results, the thin diaphragm wall, is selected for the FGS. Further technical

assumptions are for the same as those used in the reference scenario. Fig. 8a demonstrates that

for a short time period of 10 years, expected long-term advantages for the FGS are not yet

accentuated. Extending the planning phase beyond 10 years, however, increasingly favors the

FGS with respect to the depletion of energy resources. This is especially caused by the higher

GAC consumption for PTS and the permanent drain for pumping groundwater. Aside from this,

the calculation of inhabitant equivalents (Ieq.) here means an averaging of all impacts over the

entire planning period. The longer the time frame assumed, the less the annually averaged impact

of the dominating initially caused environmental burdens. After 20 years, the overall potential of

the FGS to defer ecological damages is lower than that of the respective PTS (Fig. 8b). The total

human health impacts for the PTS accumulate over time, and, as can be seen in the results for 50

years (Fig. 8c), approach the impacts of the FGS which are predominantly associated with the

construction of the steel casings.

A possible point of discussion is the assumption of equal clean-up (or operation) time for the

two completely different remediation technologies. In fact, the time required to finally achieve

the remedial objectives, such as sufficient decontamination, may vary among the different

technologies. For example, it could be anticipated that the higher groundwater flowrate

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199192associated with the application of a PTS, will speed up the remediation, through a higher

contaminant extraction rate. However, since todays knowledge is simply insufficient to guess

the exact remediation progress of such applications in the future, forecasting long-term changes

is highly speculative. Therefore, a detailed quantitative examination of the influence of

contaminant concentration variations will not be addressed here. The concept of LCA based on

bplanning periodsQ instead of potential remediation times particularly makes sense, because, afterdecades of operation, there is a high probability that innovative approaches will supersede

todays standard technologies. Consequently, not remediation time but technical evolution will

be the determining factor for the bgraveQ of the products and services considered within theLCA, particularly bearing in mind the long time frame involved.

18. Influence of GAC consumption

There is a high uncertainty in the prediction of GAC consumption over time. If it is suspected

that the remediation of the aquifer can be achieved within the planning period, e.g. within 30

years for the reference scenario, then assuming a constant contaminant concentration in the

treated groundwater over time does not seem realistic. This would mean that the transfer of

contaminants into the groundwater stops abruptly after 30 years. However, a common

observation is that the concentration in the treated groundwater gradually decreases. If

concentrations become lower as a consequence, the demand on GAC will simultaneously

decline. In view of this, the assumption of a constant GAC consumption rate during the

remediation appears rather conservative.

Since the estimated demand on GAC is assumed to be higher for the PTS than for the FGS, a

decline of the GAC mass per m3 groundwater treated would especially favour PTS. The essential

role of GAC, especially for the categories DER, GWP and POCP, is depicted in Fig. 5a and b

and the influence of any GAC cutback can be appraised. Another issue is that, as GAC lifetime

is not often determined exclusively by its sorption capacity and the prevailing contaminant

concentrations, this matter could be a disadvantage for the FGS. Long-term applications of GAC

bear the risk of bio-fouling or mineral clogging, causing a reduced performance of the reactors

and thus necessitating an early exchange of treatment material (Kraft and Grathwohl 2002).

Since, due to the higher volume of GAC per fill, the lifetime of in-situ-reactors significantly

exceeds the regeneration interval timeframe of an on-site reactor (here 5 years for the FGS, 1

year for the PTS), performance loss appears to be more likely for the FGS. Furthermore the PTS

is a less static implementation. Contrary to the FGS with fixed gate configurations, GAC vessels

can be dynamically replaced by others of a different size in order to control the length of the

regeneration period.

There is a significant influence of the technology-specific GAC on the impact indicator values,

which cannot be precisely estimated and could be further addressed by more candidate scenarios.

However, the number of possible alternatives is endless. Instead, we ask what the hypothetical

PTSs GAC demand is necessary to balance the overall impacts of both technologies? The FGS is

configured according to the Karlsruhe site and, as described for the reference scenario, with GAC

exchanges every 5 years at a constant consumption rate. Additionally, the alternative FGS using

different wall types are inspected. The configuration of the PTS agrees with the reference case

although the GAC volume is set as a variable to be adapted for different planning periods. Please

note that no uncertainty in input parameters is considered here.

Fig. 9ac show the lines of indifference, indicating the relative GAC volume per year

required for the PTS and resulting in the same potential impacts with respect to depletion of

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199 193energy resources (DER), ecological quality (EQ) and human health (HH). Above these bbreakeven trade-offsQ, the higher GAC demands of PTS delay a better rating of FGS and vice versa.

Due to the temporally increasing GAC demand and the attenuated relative importance of FGS

installation impacts, these lines have hyperbolic trends. Regardless of the wall type installed for

the funnel, if extremely large GAC demands are presumed for the PTS, long planning periods

will be necessary to obtain the same impacts on HH as for the FGS. Even the FGS with a thin

diaphragm wall, which interferes the least with human health aspects, can only compete with the

PTS at an operation time of 70 years and only if the annual relative GAC demand is assumed to

be about 250% of the FGSs demand.

With respect to other classes, the GAC threshold ratios are computed significantly lower than

those denoting potential adverse effects on human health. This is because of the higher relevance

of GAC compared to other technological elements when focusing on DER and EQ, which is

especially influenced by the high resource and energy demand for production, supply and

regeneration of GAC. Therefore, with respect to DER, a lower consumption of GAC for the FGS

is computed that leads to the same overall emissions as for PTS. We observe a steeper decline of

Fig. 9. Comparison of PTS and FGS with respect to potential depletion of resources (DER), the expected encroachment

of ecological quality (EQ) and the hazardous effects on human health (HH), assuming that the GAC demand for PTS

(GACPTS) is not known. Trade-off curves depict the relative consumption rate GACPTS/GACFGS where the total class-

specific impacts are equal, given a fixed estimated consumption of GACFGS=40 t/year for the FGS. Above the curves,

FGS are beneficial, below, PTS.

P. Bayer, M. Finkel / Journal of Contaminant Hydrology 83 (2006) 171199194the trade-offs with respect to EQ versus DER so that intersections occur after 40 years or more.

Fig. 9ac depict the GACPTS/GACFGS ratios assumed for the reference case (180%) and the

equality line, at GACPTS/GACFGS=100%, as dashed lines. Anticipating that the extraction rate

for the PTS (18 l/s, FGS: 10 l/s) is not underestimated, an 80% higher annual GAC demand for

the PTS may serve as an upper limit. This is because the contaminant concentrations can be

expected to be equal to or, due to dilution, lower than those given for the FGS. Further

arguments for this postulation are the above-mentioned adaptability and the better control of the

on-site treatment unit. Moreover, it is doubtful if, in all gates, the maximum sorption capacity of

the GAC can be reachedanother potential disadvantage for FGS.

As can be seen for all three wall types, human health related emissions of the FGS always

surmount those for PTS under realistic assumptions for the planning period and PTSs GAC

demand. This relationship is especially controlled by the usage of steel. Please note that for the

scenarios discussed here, credits from steel recycling are already considered. With respect to

DER, the 180% threshold can already be reached after 15 years of operation for all wall types. In

contrast, the break- even point for the impacts computed for the ecological quality is 1020 years

later. As already discussed above, funnels made out of thin diaphragm walls are preferable over

those constructed out of soilcementbentonite slurry walls and steel sheet piles, thus offering an

earlier break-even point. For example, assuming an 80% higher annual GAC demand for the PTS,

the depletion of energy resources is already undercut by the FGS after one decade. If the relative

annual GAC demand is lower than 180%, then the indifference trade-offs increase exponentially.

If the same annual GAC amount is consumed for both technologies (GACPTS/GACFGS=100%),

then periods of over 40 years are required to achieve lower impacts using the FGS.

19. Conclusions

There are two main contributions of this study: the set-up and the application of a LCA

framework for common long-term groundwater remediation technologies. Each step of the

LCAthe definition of functional units, system boundaries and objectives, the realisation of the

impact assessment and the comparison of technology-specific resultsnecessitated a purpose-

built approach to fit the technological key elements into the LCA framework and to obtain valid

results. The application site, the former gas manufacturing plant of the city of Karlsruhe, enabled

quantification of realistic estimates of the environmental burden caused by secondary impacts of

applying remediation technologies.

The ideal functional unit to be adopted for a LCA framework is shown to be the control and/

or remediation of a site. This is defined by complying with concentration levels, since this

reflects the perception of equal functionality according to risk assessment objectives without

examining the exact system performance. For the definition of system boundaries as well as the

spatial and temporal resolution of the environmentally relevant processes, a compromise had to

be found that most accurately outlines the adverse effects on environment and human health, but

focuses on the most crucial processes and effects of the manifold types involved. To achieve this,

key technical elements are distinguished, such as funnel installation, gate construction, pumping

devices and granular activated carbon (GAC) as the treatment material. The impact assessment

carried out then delivers the specific impacts with respect to depletion of energy resources and

burdens on ecological quality or human health. Due to the dissimilitude of the various processes

involved, the underlying inventory data for estimating product-related material demand, energy

consumption and occurring emissions is based on various literature sources, databases and

manufacturers information. As a consequence, it is hardly possible to guarantee a similar quality

for all data utilised and functional relationships assumed. Additionally, available databases are

commonly not sufficient for obtaining an exact quantitative description of all products and

services within the system boundaries from cradle-to-grave as recommended for an apposite

LCA. In view of the limited data availability and comparability, a certain degree of

bstreamliningQ was necessary, which here meant to forego a distinction of spatial and temporalcharacteristics of the processes involved.

Deficiencies in data quality and technical description were tackled by an uncertainty analysis

which yielded confidence intervals of th