at SciVerse ScienceDirect

Journal of Cleaner Production 54 (2013) 78e88

Contents lists available

Journal of Cleaner Production

journal homepage: www.elsevier .com/locate/ jc lepro

Investigating the effect of construction management strategies onproject greenhouse gas emissions using interactive simulation

Pei Tang*, Darrell Cass, Amlan MukherjeeDepartment of Civil and Environmental Engineering, Michigan Technological University, Houghton, MI 49931, United States

a r t i c l e i n f o

Article history:Received 30 July 2012Received in revised form17 March 2013Accepted 19 March 2013Available online 11 April 2013

Keywords:Project greenhouse gasConstruction management strategiesInteractive simulationEmpirical analysis

* Corresponding author. Tel.: þ1 906 4871952; fax:E-mail address: ptang@mtu.edu (P. Tang).

0959-6526/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.jclepro.2013.03.046

a b s t r a c t

The challenges posed by global climate change are motivating the investigation of strategies that canreduce the life cycle greenhouse gas (GHG) emissions of products and processes. While new constructionmaterials and technologies have received significant attention, there has been a limited emphasis onunderstanding how construction processes can be best managed to control GHG emissions. Unexpecteddisruptive events tend to adversely affect construction costs and delay project completion. They also tendto increase project GHG emissions. The objective of this paper is to investigate ways in which projectGHG emissions can be controlled by appropriately managing disruptive events. First, an empiricalanalysis of a specific highway construction project is used to illustrate the impact of unexpected scheduledelays in increasing project GHG emissions. Next, a simulation based method is introduced to assess theeffectiveness of alternative project management strategies in controlling GHG emissions. It demonstratesthat appropriately selected strategies can reduce project GHG emissions without increasing the con-tractor’s financial burden or causing project schedule delays. The contribution of this paper is that itexplicitly considers project emissions, in addition to cost and project duration, in developing projectmanagement strategies. Practical application of the method discussed in this paper will help construc-tion firms reduce their project emissions through strategic project management, and without significantinvestment in new technology. In effect, this paper lays the foundation for best practices in constructionmanagement that will optimize project cost and duration, while minimizing GHG emissions.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

According to the U.S. Environmental Protection Agency (EPA),the construction sector accounts for 131 million metric tons of CO2equivalent (EPA, 2009). GHG emissions from construction andrehabilitation of highway infrastructure make up 13.22% of theemissions in the construction sector (EPA, 2006a). The challengesposed by global climate change are motivating the investigation ofstrategies that reduce the life cycle GHG emissions of products,processes, and services. While novel construction materials andtechnologies have received significant attention (Gambatese andRajendran, 2005; Horvath and Hendrickson, 1998; Muga et al.,2009; Zapata and Gambatese, 2005), there has been a limitedemphasis on studying the construction phase to understand ifbetter management of construction processes can lead to reductionof project GHG emissions.

Close monitoring of a few highway construction projects showsthat poor management during the construction phase results in

þ1 906 4872943.

All rights reserved.

higher project GHG emissions. A detailed case study presented inthis paper illustrates the differences observed between the as-planned and the actual as-built project GHG emissions. These ob-servations reveals that unexpected interruptions and delays in-crease the total project GHG emissions. Often the underlying causeis the increased material and equipment used on-site, as comparedto the planned use. Guggemos and Horvath (2006) pointed out thatequipment use accounts for 50% of most types of emissions andenergy use of construction processes. In addition, EPA (2006b) re-ported that in 2005, construction equipment generated roughly32% of all land-based non-road NOX emissions and more than 37%of land-based Particulate Matter (PM10). In general, non-roadequipment has higher emissions than heavy-duty highway vehi-cles and automobiles (EPA, 2006b). Ahn and Lee (2013) investigatedthe environment improvement by controlling equipment efficiencyin construction operations. Ozcan-Deniz et al. (2012) used geneticalgorithms to optimize construction operations with an objectivefunction of project time, cost, and environmental impacts. But theydid not investigate the relationship between managing projecttime, cost, and environmental impacts.

In light of these observations and findings, this paper aims toinvestigate the relationship between construction management

P. Tang et al. / Journal of Cleaner Production 54 (2013) 78e88 79

strategies and increased project GHG emissions. A constructionmanagement strategy is defined as a sequence of decisions tomanage the project, given the project priorities and outcomes. It ishypothesized that the success of achieving project objectives andpriorities is dependent on understanding ways of developingcoherent management decision sequences. Appropriately selectedmanagement strategies can better manage equipment usage. Inturn, it is likely that such strategies can reduce project GHG emis-sions. However, as with cost and duration, there may be a trade-offbetween cost, duration, and GHG emissions given different man-agement strategies.

Therefore, this paper introduces and implements a simulationbased method that can be used to experimentally investigate therelationships between project cost, duration, and GHG emissions bytesting different management strategies. The first part of this paperpresents an empirical analysis of a highway construction project toillustrate the impacts of schedule delays on project GHG emissions.The second part of the paper uses the same project in an experi-mental simulation platform to analyze the impact of differentmanagement strategies on project cost, duration and emissions.Project GHG emissions for the planned project schedule (referred toas ‘as-planned emissions’) and each of the simulated outcomes(referred to as ‘as-simulated emissions’) are estimated andcompared to the actual project GHG emissions (referred to as ‘as-built emissions’basedonobserveddata). All emissions are estimatedbased on as-planned, as-simulated and as-built material andequipment use on the project site. This paper does not describedetailed emission calculations. However, the underlyingmethod canbe found in co-authors’ previous work (Cass and Mukherjee, 2011).

Traditionally, construction project planning considers trade-offsbetween project cost and duration. Project GHG emissions shouldbecome a third objective that needs to be explicitly considered, aswell. The primary contribution of this paper is that it introducesproject GHG emissions as a third leg in the time-cost trade-offproblem and investigates the relationships between project dura-tion, cost, and GHG emissions. The proposed simulation method isexpected to support project planning by identifying ways to opti-mize cost and schedule performance, while minimizing projectGHG emissions.

2. Background

State agencies are increasingly implementing regulations thatencourage stakeholders in the construction industry to reduceproject GHG emissions. At the state level, Departments of Trans-portation have launched a movement toward addressing sustain-ability through the life cycle of a pavement. Some of these effortsare focused on the monitoring and reduction of project GHGemissions in construction operations. For example, California haslegislatively mandated the reduction of GHG emissions across allsectors (CEPA, 2006).

There are different ways of assessing and reducing GHG emis-sions of highway construction operations. The life cycle assessment(LCA) perspective supports the choice of products and processesthat reduce GHG emissions during the different life cycle phases,namely raw material mining, production and manufacturing, con-struction, service and end-of-life (Santero et al., 2011). Currentinvestigation of GHG emissions emphasizes the selection ofresource on site (Cass andMukherjee, 2011; EPA, 2006b; Guggemosand Horvath, 2006). This paper considers emphasizing constructionstrategy management. In preliminary research, Cass et al. (2011)has shown that there is a likely relationship between project de-lays and increased GHG emissions.

While decisions regarding use of different pavement materialtypes are often made at the agency level, decisions to reduce GHG

emissions during the construction phase arewithin the contractors’control. Sometimes such decisions place a financial burden on thecontractors. For example, equipment larger than 175 HPmade priorto 1996 tend to produce more GHG emissions than recent models(Guggemos and Horvath, 2006). This may require a contractor toconsider a potentially expensive fleet update to achieve lowerproject emissions. In contrast, this study presents a method thatadvocates inexpensive improvements to planning and manage-ment to reduce emissions.

Methodologically speaking, the primary challenge of thisresearch lies in directly observing the impact of different man-agement strategies on project performance, for the same projectand given the same conditions. Often the impacts of a particularstrategy can be undermined by the occurrence of an unexpecteddisruptive event external to the project - such as bad weather or achange order. The timing of such external events plays a crucial rolein deciding the ultimate fate of a strategy. This points to the use ofstatistical and simulation based methods that allow the assessmentof alternative strategies. Zhou et al. (2012) proposed analytic andsimulation models to identify optimal or near-optimal green pro-duction strategies. Changbum Ahn et al. (2010) used discrete-eventsimulation to plan construction operations by comparing GHGemissions of alternative operations. However, the method did notconsider the decision-makers’ responses to contingencies duringthe construction process.

This research uses a general purpose interactive simulationplatform, the Interactive Construction Decision Making Aid(ICDMA) (Anderson et al., 2009; Rojas and Mukherjee, 2006a). Itcan be used as a test bed for multiple simulations, that are rununder varying conditions and management strategies, for a givenconstruction project. It allows decision-makers to respond toproject contingencies by (re)allocating resources during the simu-lation process. Previous research in ICDMA has establishedanalytical techniques to assess strategies’ performance in cost andschedule management (Tang et al., 2010a, b). This paper appliessimilar assessment techniques to assess the impact of alternativestrategies on project GHG emissions.

The second methodological challenge is to validate the simula-tion method. The as-built history is only a particular instance of aproject realization in reality. Comparing this single realization tothe distribution of project histories generated from the simulatedenvironment is not a true validation. However, it does provide areality check on the reasonableness of the simulation outcomes andis a step in the right direction (Martinez, 2010). It is expected thatthe true validation of such methods lies in longitudinal simulationsacross multiple projects in the long term, which is our future workand not addressed in this paper. For the purpose of this paper, wecompared the as-simulated outcomes against the actual as-builtoutcome to check the reasonableness of the simulation outcomes.The case study used is a real highway construction project, whichwas closely observed and documented. The next section describesthe empirical study of the highway construction project inquestion.

3. Empirical analysis

The empirical analysis involved a ten-mile concrete pavementre-construction project in Southeast Michigan. It studied the re-construction of the East Bound section of a major interstate high-way in 2009. This section illustrates the gaps between as-plannedand as-built project GHG emissions due to project delays. First, itexplains how as-planned and as-built project data is collected.Second, it calculates as-planned and as-built project GHG emis-sions. Third, it identifies and analyzes the differences.

P. Tang et al. / Journal of Cleaner Production 54 (2013) 78e8880

3.1. Data collection

3.1.1. As-planned data collectionThe as-planned project data was collected from the progress

schedule (MDOT Form 1130) and the electronic proposal docu-ments of the project. The progress schedule is a document sub-mitted to the Michigan Department of Transportation (MDOT) bythe contractor before the project starts. The form outlines con-struction activities (Table 1) along with proposed starting and enddates for each activity. When preparing the resource loaded as-planned schedule, the quantities in the contractor’s bid wereused. Activities defining the actual construction of the highwaywere identified. These activities were already associated with adivision of work and section number as defined in the MDOTStandard Specifications for Construction (MDOT, 2003). Associatedpay-items, which were in the electronic proposal of the project,were identified for each activity. The bid tab quantities, for eachpay-item, represented the entire project, not just the mainline.Therefore, the ratio of as-built mainline quantities to that of thetotal quantities was applied to the bid tab quantities to calculate theas-planned quantities for each of the primary activities.

RS Means 2009 cost data (RS Means Cost Works, 2009) was usedto estimate the cost of the project so that a price could be ascribed toeach activity. Labor productivity was estimated using cost informa-tion from labor crews associatedwith thematerials assigned to eachpay-item. The number of the labor crews and their contributions toeach activity were determined by grouping similar labor crews. RS-Means represents the average prices of resources for the industry.Project budget estimates using RS-Means prices are typically higherthan contractors’ actual cost. Though the unit price data could not becollected from the contractor, the expected activity durations, actualactivity durations, and the quantity of work to be completed wereavailable. Therefore, it is expected that the project price wasreasonably estimated and, for the purpose of this analysis provides auseful benchmark to work with. The collected as-planned data sup-ported the development of a resource loaded as-planned schedule.

3.1.2. Actual as-built data collectionMDOT requires the use of software called FieldManagerTM

created by InfoTech Inc on all their construction and rehabilitationcontracts. Inspectors (on behalf of MDOT) use the software to re-cord daily general site information, contractor personnel andequipment, and postings of material quantities. The software gen-erates an Inspector’s Daily Report (IDR) that stores all this

Table 1Activities’ information and constraints between them.

No. Activity description Duration Precedenceactivities

Spat

1 Strip topsoil 10 days n/a2 Remove concrete pavement 30 days 1 0.53 Grade subbase 26 days 2 n/a4 Install drainage 18 days 2,3 0.5

1 to5 Place open graded drainage course

(OGDC) mainline18 days 2,3,4 1 m

6 Pave east bound mainline 32 days 5 1 m7 Place OGDC ramps and gaps 8 days 4,5,6 3 to8 Pave east bound gaps and ramps 9 days 7 n/a9 Place gravel shoulder 4 days 4 2 to10 Slope grading and restoration

east bound26 days 9 0.5

east11 Stripe to open pavement east

bound3 days 9 10 m

12 Relocate barrier wall 10 days 10 10 m13 Re-stripe west bound 3 days 12 n/a14 All lanes open 1 days 12,13 n/a

information. The following three fields of the IDRs were used toinvestigate the actual as-built schedule.

1) General Site Information: This field defines the context withinwhich the construction operations for the day were conducted. Itincludes the days and time the work was performed, weatherconditions on-site, general observations about site conditions, andsite-specific location of the work being performed (including sta-tion information). Disruptive events, like bad weather, thatadversely affected the project schedule were also reported.

2)ContractorEquipment: Thecontractorpersonnelandequipmentinputs of the IDRswere critical to quantifying project GHGemissions.The types, hours of use, and make of construction equipment usedon-site significantly impact a project’s total GHG emissions. Equip-ment used throughout the project was recorded using this field.

3) Material Posting: IDRs tracked progress on each pay-item asspecified in the construction contract. It recorded the location,station information and quantities of materials associated witheach activity. This data was used to develop an actual as-built re-cord of procured and installed pay-items. Using as-built quantitiesto calculate life cycle impacts and GHG emissions is significantlymore representative of project impacts compared to similar cal-culations done with estimated quantities.

The information in the IDRs was organized in a database andqueried to extract as-built data associated with daily pay-item in-formation. The daily material use for each activity was establishedby associating pay-items with activities. The quantities of materialinstalled, equipment used, station locations, and the inspectors’observations were queried for each day.

Besides investigating IDRs, the researchers visited the con-struction site and corresponded with the contractor to ensure theaccuracy of the data collected through FieldManagerTM. Equipmentwas checked to ensure that equipment posted in the IDRs matchedequipment actually used on-site. The fuel usage and fleet data fromthe contractor was also collected to identify equipment specifics.All these collected items validated and supported the data collectedthrough FieldManagerTM and direct reports from contractors.Through correspondence with the primary contractor of this proj-ect, the as-planned schedule and the actual activity constraints inthe construction process were obtained.

3.2. Project GHG emission estimation

The GHG emissions are estimated for installed materials and theequipment used for the project. The co-authors present a detailed

ial distance between activities

to 1 miles between stripping topsoil and concrete pavement removal

to 1 miles between installing drainage and concrete pavement removal;3 miles between installing drainage and grade subbaseile between grade subbase and placing OGDC mainline

ile between placing OGDC mainline and paving east bound mainline5 miles between paving east bound mainline and placing OGDC ramps and gaps

3 miles between paving east bound gaps and ramps and placing gravel shouldermiles between placing gravel shoulder and slope grading and restorationboundiles between placing gravel shoulder and striping to open pavement east bound

iles between striping to open pavement east bound and relocating barrier wall

Table 2Hourly GHG emission rate for equipment using diesel.

No. Equipment Loadfactor

Avg. HP GHG emissionrate (lbs CO2/hr)

1 Aerial lifts 0.46 60.49 9.502 Air compressors 0.48 105.67 17.313 Bore/drill rigs 0.75 291.19 74.554 Cement and mortar mixers 0.56 10.32 1.975 Concrete/industrial saws 0.73 18.61 4.646 Cranes 0.43 399.10 58.587 Crawler tractors 0.64 146.89 32.098 Crushing/proc. equip. 0.78 142.34 37.909 Excavators 0.57 168.08 32.7010 Forklifts 0.30 144.59 14.8111 Generator sets 0.74 549.20 138.7312 Graders 0.61 173.71 36.1713 Off-highway tractors 0.65 266.98 59.2414 Off-highway trucks 0.57 478.94 93.1815 Other construction equip. 0.62 74.69 15.8116 Other general industrial equip. 0.51 238.06 41.4417 Other material handling equip. 0.59 190.84 38.4318 Pavers 0.62 100.23 21.2119 Paving equip. 0.53 103.70 18.7620 Plate compactors 0.43 8.00 1.1721 Pressure washers 0.60 0.91 0.1922 Pumps 0.74 53.46 13.5023 Rollers 0.56 95.40 18.2424 Rough terrain forklifts 0.60 93.41 19.1325 Rubber tired dozers 0.59 357.06 71.9126 Rubber tired loaders 0.54 157.00 28.9427 Scrapers 0.72 312.50 76.8028 Signal boards 0.78 20.16 5.3729 Skid steer loaders 0.55 43.87 8.2430 Surfacing equip. 0.45 361.88 55.5931 Sweepers/scrubbers 0.68 91.14 21.1532 Tractors/loaders/backhoes 0.55 107.98 20.2733 Trenchers 0.75 62.76 16.0734 Welders 0.45 45.43 6.9835 Work trucks, haul trucks semis 0.75 250.00 64.0036 Generator 0.75 20.00 5.12

Fig. 1. As-planned v.s. as-built progress schedule.

P. Tang et al. / Journal of Cleaner Production 54 (2013) 78e88 81

analysis and illustration of how GHG emissions are calculated forpavement construction using LCA tools (Cass andMukherjee, 2011).The Economic Input OutputeLife Cycle Assessment (EIO-LCA) andSimaPro (using the Eco Invent database) were used to estimate theimpacts of materials through the life cycle stages of extraction/mining, manufacturing, and transportation. The data collectedthrough FieldManagerTM was used to develop material, and fuelinventories, which in turn were used as an input to the LCA tools.This method is not described in this section to avoid repetition. Thedetailed inventories can be found in PE-2 (Mukherjee, 2011).

When assessing equipment GHG emissions, the as-planned andas-built equipment usage per working day is required. The make,model, type, and horsepower characteristics of each equipmenttype were identified in the fleet information provided by thecontractor. Using Equation (1), the hourly GHG emissions wereestimated for each equipment type that uses diesel fuel. TheEquation (1) is an adapted equation from EPA (2010).

Hourly Equipment GHG Emission Rate ¼ Ot*Lf *HP*CF*ε (1)

Where Ot ¼ Operating time factor, which is relative to how long theequipment operates; Lf ¼ Average load factor, which is relative to atwhat horsepower capacity the equipment operates; HP ¼ Averagerated horsepower; CF ¼ Fuel consumption rate (Gal/(HP*hr)), andε ¼ GHG emission rate (lbsCO2/Gal). The following assumptionswere made:

� Operating time factor Ot ¼ 45 min/hr (0.75)� Fuel consumption rate CF ¼ 0.04 (Gal/(HP*hr)) (Peurifoy andOberlender, 2002)

� GHG emission rate ε ¼ 22 (lbsCO2/Gal) (EPA, 2005)

Average load factor and rated horsepower for equipment wereobtained from Roadway Construction Emissions Model (SCAQMD,2009). Table 2 shows the hourly GHG emission rate for eachequipment type.

3.3. Description of actual as-built progress and uncertainties

The scope of this analysis is to investigate project GHG emis-sions associated with the highway re-construction process, so ac-tivities are chosen that are representative of typical highwayconstruction projects. These activities are referred to as primaryactivities. Traffic control activities were excluded because they mayvary significantly from project to project. Pavement removal,earthwork and paving operations are considered as primary ac-tivities. For each primary activity, a controlling pay-item wasidentified to represent the activity. They were:

� Primary Activity: Remove Concrete Pavement, Controlling Pay-item: Pavement Removal

� Primary Activity: Grade Subbase, Controlling Pay-item: StationGrading

� Primary Activity: Install Drainage, Controlling Pay-item:Underdrain Pipe

� Primary Activity: Place Base Material, Controlling Pay-item:Geotextile Separator

� Primary Activity: Pave Mainline, Controlling Pay-item: Non-reinforced Concrete

Each activity was assigned a controlling equipment that wasused to complete the controlling pay-item, as follows:

� Remove Concrete Pavement: Pavement Breaker� Grade Subbase: Grader

� Install Drainage: Trencher� Place Base Material: No equipment was required by the con-trolling item

� Pave Mainline: Concrete Paver

Fig. 1 shows the as-planned and actual as-built completion ratesfor each activity. The x-axis represents the time, and the y-axisrepresents the cumulative completion percentages. Due to

Table 3Quantity comparison of consumed controlling pay-item.

Primary activity Controlling item Unit AsPlanned Qty AsBuilt Qty % change

Remove concrete pavement Pavement removal Syd 249,066 185,432 �25.55Grade subbase Station grading Syd 449 519 15.75Install drainage Underdrain pipe Ft 110,008 107,945 �1.87Place base material Geotextile separator Syd 213,236 217,750 2.12Pave mainline shoulder Non-reinforced concrete Syd 217,359 229,876 5.76

P. Tang et al. / Journal of Cleaner Production 54 (2013) 78e8882

unexpected disruption on site, project activities were performeddifferently from the as-planned schedule. Pavement removal ac-tivity started later than as planned. Delays in pavement removalactivity did not affect the grade subbase activity starting on time.The rest of the activities started earlier than planned, but mainlinepaving activity took five times longer than the as-planned tocomplete the job. IDRs from FieldManagerTM showed that the ac-tivities were interrupted by: (1) seasonal rain, which added to theflooding on-site and brought all operations to a standstill, (2)equipment breakdown, worker illness or accidents, whichdecreased activities’ productivity, (3) the concrete test failure,which interrupted the progress because extra test(s) was required,(4) traffic accidents on the construction site, (5) rework in pave-ment removal between the fifth and the sixth mile points, causedby the ground water flooding due to agitation of the soil during thepavement removal and the presence of on-site heavy equipment.Besides the disruptive events, the discontinuous gaps along theproject also delayed the project. For example, in Fig. 1, the mainlinepaving activity was halted for two weeks on August 20, 2009 andSeptember 23, 2009.

3.4. Gaps between as-planned and actual as-built GHG emissions

Compared to the as-planned quantities of controlling pay-item(Table 3), grading subbase activity, placing base material activity,and paving mainline and shoulder activity actually consumedrespectively 15.75%, 2.12%, and 5.76% more pay-items. Removingexisting concrete pavement activity and installing drainage activityused 25.56% and 1.87% fewer pay-items than planned. Paving themainline and shoulder activity produced 800 metric tons (mt) ofextra emissions (Table 4), which is 5.88% more than planned.Installing drainage activity produced 2% fewer emissions thanplanned and placing base material activity produced 2.11% morethan planned. Material that cannot be stored indefinitely shouldpartly account for the extra material usage. For example, concretewas disposed if it was not used on the day it was produced.

Except the grading subbase activity, high exceedance percent-ages were found when comparing the as-built equipment emis-sions to the as-planned (Table 5). Pavement breaker produced 60%extra emissions due to rework and non-rework disruptions. Thetrencher and the concrete paver respectively produced 57.14%, and85.71% more emissions mainly because of non-rework disruptionsidentified as bad weather, equipment breakdown, worker sickness,on-site accidents, and concrete test failure.

Rework has weaker influences than non-rework relateddisruptive events in this project because rework only influenced

Table 4Comparison of GHG emissions produced by controlling pay-item.

Primary Controlling

Remove concrete pavement Pavement removalGrade subbase Station gradingInstall drainage Underdrain pipePlace base material Geotextile separatorPave mainline shoulder Non-reinforced concrete

the pavement removal activity. Non-rework related disruptiveevents assumed more responsibilities for the project delays andextra GHG emissions. When considering the as-built GHG emis-sions from all the equipment andmaterial instead of the controllingones, a larger amount of project emissions was found due to thenon-rework related disruptive events. Therefore, the pertinentquestions raised are: (1) is it possible to reduce project GHGemissions through better construction management strategies, and(2) what are the relationships between project total cost, comple-tion date, and GHG emissions? To answer these questions, the nextsection introduces an interactive simulation experiment, fromwhich decision-makers can explore and compare the performanceof alternative construction management strategies.

4. Simulation experiment

The experiment is performed on the interactive simulationplatform of ICDMA (Anderson et al., 2009). ICDMA simulates aconstruction project based on its resource loaded as-plannedschedule and project environment. The resource loaded as-planned schedule provides a baseline to complete the project,and the project environment is defined by the possible disruptiveevents that deviate the project from the as-planned schedule.During a simulation run, decision-makers are presented withrandom external events, thus allowing them to respond to dis-ruptions. The decision-makers apply a specific strategy to managethe contingencies by (re)allocating resources. ICDMA takes theresponse and updates the project. The consequences of the de-cisions result in new scenarios for decision-makers to respond to.This process continues until the completion of the simulatedproject. The simulation experiment is expected to identify appro-priate management strategies to control project GHG emissions.

4.1. Simulation project setup

To set up the highway reconstruction project in ICDMA, aresource loaded as-planned schedule and construction environ-ment are required. The following process is used.

� Input general information for material, labor and equipment.This includes material description, unit cost, and material in-formation such as e whether the material can be storedindefinitely or not.

� Input material, labor and equipment usage information foreach activity. This includes the set-up of labor crews, involving

AsPlanned mtCO2eq AsBuilt mtCO2eq %

No consumption of virgin materialsNo consumption of virgin materials45 44.1 �2379 387 2.1113,600 14,400 5.88

Table 5Comparison of GHG emissions produced by controlling equipment.

Primary activity Contr. equip. AsPlanned AsBuilt AsPlanned AsBuilt %

# of working days # of working days GHG emissions mtCO2eq GHG emissions mtCO2eq Change

Remove concrete Pavement Pavement Breaker 15 24 5.66 9.06 60.00Grade subbase Grader 19 8 14.87 6.26 �57.89Install drainage Trencher 14 22 5.29 8.31 57.14Place base material Geotextile separator placed by 4-person crew, no equipmentPave mainline &shoulder Concrete 14 26 11.63 21.60 85.71

P. Tang et al. / Journal of Cleaner Production 54 (2013) 78e88 83

the input of crew descriptions, the quantities and types of laborand equipment, and the quantities of material for each activity.

� Input the constraints between the activities. This includesprecedence relationships and activity/resource constraintsdriving the schedule.

� Set up risk environment by defining disruptive events, associ-ated probabilities, and their consequences to the project(Table 6). The disruptive events were recorded in IDRs. Theirprobabilities were estimated by dividing occurrence fre-quencies by the actual project duration. Bad weather, equip-ment failure, worker illness, and concrete test failure werefound to be the three most influential events that should beresponsible for project delays.

Spatial constraints between activities, such as the length ofhighway that must separate equipment associated with any twoactivities, were measured in distances provided by the primarycontractor (Table 1). They were converted to temporal equivalentsbecause ICDMA uses temporal constraints. For example, the con-struction manager required a distance of three miles between thesub-base grading activity and the drainage installation activity,depending on the frequency of drainage crossings along themainline. Sub-base grading operation has a productivity rate of0.53 miles/day (¼ 10.14 miles/19 days). Therefore, it was decidedthat the drainage installation activity would start six days (¼ 3miles/0.53 (miles/day)) after the beginning of the sub-base gradingactivity. Activities of removing concrete pavement, grading the sub-base, installing the drainage, placing OGDC along the mainline, andpaving the east bound mainline were divided into three segmentsto better represent the constraints (Hegazy and Menesi, 2010). Thetotal as-planned duration was estimated to be 106 working daysand the critical activities are marked in the shaded boxes (Fig. 3).Fig. 2 showed the interface of the highway reconstruction project inICDMA.

The general information for material, labor and equipment, andtheir usage for each activity were obtained in the as-planned data.

Table 6Project environment.

Disruptive event Precondition Postcondition Probability

Bad weather N/A The productivity rates of eachactivity reduce by 50% inbad weather.

0.20

Equipmentfailure orworker sick

N/A Random labor(s) is(are) sickor equipment break(s) downwhen the event occurs. Impactsof the event on the activityproduction rate is calculatedbased on the weight of laboror equipment in the specificlabor crew.

0.12

Concretetestingfailure

Pavingconcrete isready.

Productivity rate of pavingactivity reduces to zero forone day waiting forclearance of new test.

0.05

The disruptive events, associated probabilities, their consequences,and activity constraints were acquired from the as-built data.

4.2. Construction management strategy development

Once the simulation was set up, multiple experiments wereconducted to test and compare the performance of different stra-tegies. In each experiment, a decision-maker ran the simulatedproject using a specific management strategy. Each strategy reflectsa set of priorities governing project cost and duration. For illus-tration purposes, this research considers three strategies, a controlstrategy, a passive management strategy, and an aggressive man-agement strategy. Passive management is based on ad-hocresponse to unexpected events. Aggressive management aims toanticipate the future and prepare contingency plans that minimizeadverse consequences from unexpected disruption to the projectschedule. Strictly speaking, the control strategy is a special case ofpassive management strategies. It manages the schedule by takingthe minimum number of actions in dealing with interruptions andis used as a baseline strategy to contrast the impacts of other stra-tegies. Each decision strategy consists of policies that determine theordering, and (re)allocation of material, labor and equipment re-sources. Different allocation policies directly influence each activ-ity’s productivity rate. For example, in case the only paver on sitebreaks down, the productivity rate of paving activity will reduce tozero when the event occurs. Other activities’ productivity ratesremain unchanged. When resources policies are applied, a cost ofusing the resource(s) occur and is calculated correspondingly. Tounderstand how ICDMA functions and deals with events such asdisruptions and decisions, please refer to co-author’s previous work(Anderson et al., 2009; Dossick et al., 2010; Rojas and Mukherjee,2003, 2005, 2006). The strategies considered are as follows.

� Control Strategy: No actions are taken when the project isfalling behind. Resource allocation policies: (a) labor crewpolicy: no extra workers are replaced in any case; (b) equip-ment policy: equipment is fixed the next day if it breaks downand never hire extra equipment; (c) space policy: critical ac-tivities are prioritized when allocating space on site.

� Catch Up Strategy: It belongs to passive management strate-gies. The decision-maker who uses this strategy passively ap-plies resource allocation policies to catch up with the scheduleevery time the project is three days behind the as-plannedschedule. When approaching the end of the project, theCatch Up Strategy reduces the tolerance of schedule delay toone day for timely completion. Resource allocation policies: (a)labor policy: extra workers are hired and replaced in cases ofillnesses and project delay; (b) equipment policy: equipment isfixed by the mechanics immediately and extra equipment isused in case of schedule delay; (c) space policy: critical activ-ities are prioritized when allocating space on site.

� Crash Strategy: It belongs to aggressive management strate-gies. The decision-maker who uses this strategy assesses future

Fig. 2. Interface of simulated highway reconstruction project in ICDMA.

P. Tang et al. / Journal of Cleaner Production 54 (2013) 78e8884

risks and applies resource allocation policies to stay three daysahead of the as-planned schedule. When approaching the endof the project, the Crash Strategy reduces the goal of stayingthree days ahead of the as-planned schedule to one day fortimely completion. Resource allocation policies: (a) labor

Fig. 3. Activity

policy: extra workers are hired and replaced in cases of ill-nesses or to expedite the schedule; (b) equipment policy:equipment is fixed by the mechanics immediately and extraequipment is used to expedite the schedule; (c) space policy:critical activities are prioritized when allocating space on site.

diagram.

P. Tang et al. / Journal of Cleaner Production 54 (2013) 78e88 85

4.3. Simulation experiment and data collection

Each of the three strategies was implemented to complete thesimulation project for thirty five runs in order to meet the min-imum requirement of statistical analysis. For the sake of unifor-mity and given the experimental nature of this research, a singledecision-maker (one of the authors, in this case) ran all thesimulations. During each simulation run, the following data wascollected: (1) total cost and duration at the completion of theproject, and (2) daily material, equipment, and labor usage. TheGHG emissions were calculated for equipment by multiplying theequipment hours by their hourly GHG emission rates in Table 2.Because each simulation run is independent, it was safelyassumed that the results are independent and identicallydistributed for each strategy, when the number of experimentswas large.

4.4. Comparing alternative construction management strategies

In all, there were three data sets used for the analysis. First, asingle instance of the as-planned project schedule and the costestimated from project documents and RS-Means respectively, asdescribed in a previous section. This was the same as-plannedschedule used to set up the simulation. Second, a single instance ofthe actual as-built project schedule and final cost data collectedfrom FieldManagerTM and other construction site records, asdescribed in a previous section. Finally, the third data set wascollected from the simulation. For each of the three strategiestested in the simulation, 35 instances of the project realizationwerecollected - a total of 105 instances with a final as-simulated projectcost, schedule, and GHG emissions. In the empirical analysis, dif-ferences between the as-planned and as-built GHG emissions wereidentified. This section compares the instances of the three stra-tegies to identify preferred strategies and to recommend projectspecific management practice. A comparison between the as-builtand the as-simulated is provided as a reality check for thesimulation.

Table 7 summarizes the as-built and as-simulated total costs,durations, and GHG emissions. The as-built data showed that it took129 working days to complete the project, with a production of1683.75 mt of CO2 at a total cost of $20,277,970. In the simulationexperiment, the Control Strategy took an average 127.91 days tocomplete the project, which was comparable to the as-built dura-tion of 129 days. In addition, simulation results showed that theControl Strategy, Catch Up Strategy, and Crash Strategy respectivelyproduced 1414.58, 1427.38, and 1450.78 mt CO2. The GHG emissionquantities were comparable (within ¡8% of the as-built amount) tothe emission produced in the actual construction process(1683.75 mt CO2). While this comparison does not provide a truevalidation of the simulation, it establishes credibility for thesimulation platform. In addition, ICDMA has shown a similar

Table 7Average total cost, duration, and equipment GHG emissions of the project.

Estimatedas-planned

As-simulated As-built

Controlstrategy

Catch upstrategy

Crashstrategy

Cost ($) 28,111,603 29,993,662 29,334,228 29,298,617 20,277,970Duration

(days)106.00 127.91 106.46 105.57 129.00

GHG emis.(mt CO2)

1,167.56 1,414.58 1,427.38 1,450.78 1,683.75

application to a real steel construction project (Anderson et al.,2009).

The two-sample t-test was used to examine whether two stra-tegies produced the same GHG emissions. The Control Strategy isthe reference group for the Crash Strategy and the Catch UpStrategy to compare against. The null hypothesis and alternativehypothesis are:

� Null hypothesis H0: m(Strategy i) ¼ m(Control Strategy), where irepresents Crash Strategy or Catch Up Strategy, and m repre-sents the average project GHG emissions;

� Alternative hypothesis Ha: The means of two strategies’ GHGemissions are different;

� The significance level: a ¼ 0.05;

The test was implemented in SPSS 16.0 (Statistical Package forthe Social Sciences) (Navidi, 2006). The test variables were 35 in-stances of as-simulated GHG emissions from each of the strategies.Because the variances of the groups were unknown, the two-sample t-test was performed assuming both equal variances andunequal variances. Levene’s test was used to test the equal vari-ances between the data. When comparing the Crash Strategyagainst the Control Strategy, the significance value in the Levene’stest was 0.001 (<0.05). It indicated that the variances of projectGHG emissions from the two strategies are statistically unequal, sothe results of the t-test, which assumed unequal variances, wereused. The significance value of the t-test was 0.020 (<0.05), thusproviding evidence to reject the null hypothesis that the ControlStrategy and the Crash Strategy produced the same project GHGemissions. When comparing the Catch Up Strategy against theControl Strategy, there was not enough evidence (0.297 > 0.05) toreject the null hypothesis, indicating that the Catch Up Strategy andthe Control Strategy produced a comparable amount of projectGHG emissions. Therefore, the t-test results showed differentstrategies did produce a different amount of GHG emissions.

Similarly, the abilities of alternative strategies to control projectcost and durationwere evaluated and compared. First, the one-wayanalysis of variance (ANOVA) was used to examine whether threestrategies had the same as-simulated cost and duration. Next, thePost Hoc test identified which strategy had different mean cost orduration from others. Results from ANOVA showed that strategieshad different project costs and durations (significance values were0.000, which were less than 0.05). The Tukey HSD Post Hoc testshowed that the Control Strategy had different project costs anddurations from the Crash Strategy and the Catch Up Strategy (bothsignificance values were 0.000, which were less than 0.05). Inaddition, the Crash Strategy and the Catch Up Strategy had com-parable project cost and duration (significance values were 0.164and 0.890, which were greater than 0.05).

The statistical analysis result of strategy performance can besummarized as follows:

� Cost: Strategies did affect the project costs. The Crash Strategyand the Catch Up Strategy statistically had the same cost. TheControl Strategy had a higher cost than the Crash Strategy andthe Catch Up Strategy.

� Duration: Strategies did affect the project duration. The CrashStrategy and the Catch Up Strategy statistically had the sameduration. The Control Strategy had a longer duration than theCrash Strategy and the Catch Up Strategy.

� GHG emissions: Strategies did affect the project GHG emis-sions. The Control Strategy and the Catch Up Strategy statisti-cally produced the same amount of project GHG emissions. TheCrash Strategy produced more emissions than the ControlStrategy and the Catch Up Strategy.

Fig. 4. Contribution of GHG emissions produced by activity to the differences betweenCrash Strategy and Control Strategy.

P. Tang et al. / Journal of Cleaner Production 54 (2013) 78e8886

Compared to the Control Strategy, the Crash Strategy and theCatch Up Strategy completed the project with an average shorterduration and lower cost. However, the Catch Up Strategy producedless average project GHG emissions than the Crash Strategy. Inaddition, there are no significant differences between the emissionsof the Catch Up Strategy and the Control Strategy. Hence, the CatchUp Strategy was identified among the three as the most preferredstrategy in managing project cost, duration, and GHG emission.

5. Discussion

The empirical analysis revealed that unexpected delays andinterruptions resulted in extra project GHG emissions during theconstruction. The extra usage of equipment and materials were theunderlying causes for the increased as-built emissions. This moti-vated the investigation of the relationship between project GHGemissions and construction practices. In the simulation experi-ment, different construction management strategies were tested tocompare their performance on project cost, duration, and GHGemissions. Statistical analysis showed that strategies had signifi-cantly different influences on project cost, duration, and GHGemissions. Possible outcomes of this research are as follows:

� There exist strategies that are both cost and duration efficient.Table 7 showed that the Crash Strategy and the Catch UpStrategy maintained shorter duration (on an average of 21.90days) and lower cost (averagely $677,239) than the ControlStrategy. Compared to the Control Strategy, the Crash Strategyand the Catch Up Strategy averagely spent $75,228 more directcost but $752,468 less indirect cost. The prerequisite ofdeveloping cost and duration efficient strategy is that the

Fig. 5. Contribution of GHG emissions produced by equipment to

savings in indirect cost for reducing the schedule should behigher than the increase in direct cost.

� Cost and duration efficient strategies are not always going to beemission efficient strategies. The Crash Strategy and the Catch UpStrategyweremore effective in cost and schedulemanagementthan the Control Strategy. The Crash Strategy produced36,203.50 kg more emissions than the Control Strategy, a dif-ference of 2.55% between them. However, no statistical differ-ence was found in the GHG emissions produced between theCatch Up Strategy and the Control Strategy. The comparisonimplied that improvement in cost and duration managementdid not automatically increase emission efficiencies.

� A decision should be made on whether to focus on activities withlower net daily emission rates or to manage lower hourly emissionrate equipment. Instead of accelerating projects by crashinglowest cost critical activities, production of GHG emissions canbe managed not only at the activity level but also at theequipment level. Figs. 4 and 5 compared the contribution ofGHG emissions produced from activities and equipmentrespectively to the differences between the Crash Strategy andthe Control Strategy (36,204 kg more emissions). In Fig. 4, thex-axis represents fourteen activities in the order of net dailyemission rate. Activity 6 has the largest net daily emission rateof 15,200 kg CO2 and Activity 14 has the smallest net dailyemission rate of 127 kg CO2. The y-axis is the accumulativepercentages of GHG emissions produced by activities to thedifferences between the Crash Strategy and the Control Strat-egy. In Fig. 5, the x-axis represents equipment in order ofhourly emission rates. The attenuator truck, dump truck,flatbed truck, paint truck, vacuum truck, and water truck havethe largest hourly emission rates of 64 kg CO2 and the concretepaver has the smallest hourly emission rate of 4.6 kg CO2. They-axis is the accumulative percentages of GHG emissions pro-duced by equipment to the differences between the CrashStrategy and the Control Strategy. Fig. 4 showed the first threelargest net daily emission rate activities contributed to 79.40%of the differences. Fig. 5 showed the first 12 largest hourlyemission rate equipment contributed to 81.09% of the differ-ences. It implied that managing the first 3 largest net dailyemission rate activities was equal to managing the first 12largest hourly emission rate equipment. For this project, thecomparisons suggested managing activities is a more efficientway to control GHG emissions than managing equipment.

� Themulti-objective construction management problem shouldconsider resources associated strategies when making de-cisions. Given the nature of this highway reconstruction proj-ect, investigating strategies which prioritized crashing critical

the differences between Crash Strategy and Control Strategy.

P. Tang et al. / Journal of Cleaner Production 54 (2013) 78e88 87

activities with the lower daily project GHG emission ratesmight provide further evidence to improve strategies inreducing project GHG emissions.

� Appropriately selected strategies can reduce project GHGemissions without increasing contractor’s financial burden orcausing project schedule delays, such as the Catch Up Strategyfor this project.

� Construction projects are vulnerable to schedule delays, costand GHG emission increases because of disruptive events likesevere weather and on-site accidents. This paper suggestedthat the differences between total cost of the Control Strategyand the as-planned total cost could be used as an estimate ofpossible contingency budget. The amount is $1,882,059($29,993,662 e $28,111,603 ¼ $1,882,059) for this project.

It should be recognized that different projects may havedifferent strategy recommendations because of the different riskenvironments and resource the projects face and use. For example,whether tomanage GHG emissions at activity level or at equipmentlevel is a project specific decision that depends on the particularequipment combinations for the project. The proposed interactivesimulation based method provides a general framework to testdifferent management strategies for projects scheduled with Crit-ical Path Method. It is expected that through simulating as-plannedschedules and analyzing a large number of complete project re-alizations, we can identify optimal strategies in managing cost,schedule, and GHG emissions at the pre-construction stage. Havingsaid that, there are various limitations to apply the proposedmethod to practical projects at this time point. First, risk eventdatabase for specific types of projects, including their impacts onthe productivity rates of activities, should be built. Lots of work hasbeen done in this area (Carr and Tah, 2001; Choi et al., 2004; Zouet al., 2007; Zavadskas et al., 2010), which is expected to be com-bined in future research. Second, the definition of strategies and thestrategy evaluation system need to be significantly more robust. Asimulation based optimization method must be developed in orderto optimize strategies. Third, to examine the robustness of theproposed method, more different types of projects should be testedby performing the empirical analysis and simulation experiment.

6. Conclusion

This paper investigated the relationships between project cost,duration and GHG emissions, and established a method forexploring best construction management practices at the pre-planning stage. It demonstrates that appropriately selected strate-gies can reduce project GHG emissions without increasing thecontractor’s financial burden or causing project schedule delays. Inthe short term, this work advises the best practices involved inreducing project GHG emissions for a specific project. In the longrun, more work is needed to generalize the case study results toidentify the best construction management practices that applyacross different construction projects. Revealing the relationshipsbetween project cost, duration, and GHG emissions can expand thetheory of construction management as well.

Acknowledgment

This work is supported by the Michigan Department of Trans-portation under Contract No. #080741 and the National ScienceFoundation under Grant No. SES-0624118. Any opinions, findingsand conclusions or recommendations expressed in this material arethose of the authors and do not necessarily reflect views of Mich-igan Department of Transportation and the National ScienceFoundation.

References

Ahn, C., Lee, S., 2013. Importance of operational efficiency to achieve energy effi-ciency and exhaust emission reduction of construction operations. Journal ofConstruction Engineering and Management 139 (4), 404e413.

Ahn, C., Pan, W., Lee, S., Pentildea-Mora, F., 2010. Lessons learned from utilizingdiscrete-event simulation modeling for quantifying construction emissions inpre-planning phase. In: Proceedings of 2010 Winter Simulation Conference,pp. 3170e3176. Baltimore, MD, USA.

Anderson, G.R., Mukherjee, A., Onder, N., 2009. Traversing and querying constraintdriven temporal networks to estimate construction contingencies. Automationin Construction 18 (6), 798e813.

Carr, V., Tah, J., 2001. A fuzzy approach to construction project risk assessment andanalysis: construction project risk management system. Advances in Engi-neering Software 32 (10e11), 847e857.

Cass, D., Mukherjee, A., 2011. Calculation of greenhouse gas emissions for highwayconstruction operations using a hybrid life cycle assessment approach: a casestudy for pavement operations. Journal of Construction Engineering andManagement 1, 1015e1025.

Cass, D., Tang, P., Mukherjee, A., 2011. Managing Construction Operations to Mini-mize Greenhouse Gas Emissions. In: The 2nd International Conference onTransportation Construction Management, p. 79. Feb 7e10, Orlando, FL, USA.

CEPA, 2006. Assembly Bill 32: Global Warming Solutions Act. Available via. http://www.arb.ca.gov/cc/ab32/ab32.htm [accessed 07. 30. 11.].

Choi, H., Cho, H., Seo, J., 2004. Risk assessment methodology for undergroundconstruction projects. Journal of Construction Engineering and Management130 (2), 258e272.

Dossick, C.S., Mukherjee, A., Rojas, E.M., Tebo, C., 2010. Developing constructionmanagement events in situational simulations. Computer-aided Civil andInfrastructure Engineering 25 (3), 205e217.

EPA, 2005. Emission Facts: Average Carbon Dioxide Emissions Resulting fromGasoline and Diesel Fuel. http://www.epa.gov/oms/climate/420f05001.htm[accessed 07. 30. 11.].

EPA, 2006a. Greenhouse Gas Emissions from the U.S. Transportation Sector: 1990e2003. U.S. Environmental Protection Agency, pp. 1e2.

EPA, 2006b. Recommendations for Reducing Emissions from the Legacy DieselFleet. U.S. Environmental Protection Agency.

EPA, 2009. Potential for Reducing Greenhouse Gas Emissions in the ConstructionSector. U.S. Environmental Protection Agency, pp. 32e33.

EPA, 2010. Median Life, Annual Activity, and Load Factor Values for Nonroad EngineEmissions Modeling. U.S. Environmental Protection Agency, p. 1.

Gambatese, J., Rajendran, 2005. Sustainable roadway construction: energy con-sumption and material waste generation of roadways. In: Proceedings of theConstruction Research Congress. ASCE, San Diego, CA, pp. 102e108.

Guggemos, A.A., Horvath, A., 2006. Decision-support tool for assessing the envi-ronmental effects of constructing commercial buildings. Journal of ArchitecturalEngineering 12 (4), 187e195.

Hegazy, T., Menesi, W., 2010. Critical Path segments scheduling technique. Journalof Construction Engineering and Management 136 (10), 1078e1085.

Horvath, A., Hendrickson, C., 1998. Comparison of Environmental Implications ofAsphalt and Steel-reinforced Concrete Pavements, pp. 105e113. TransportationResearch Record, NRC, No. 1626(Environmental and Social Effects ofTransportation).

Martinez, J.C., 2010. Methodology for conducting discrete-event simulation studiesin construction engineering and management. Journal of Construction Engi-neering and Management 136 (3), 25e31.

MDOT, 2003. Standard Specifications for Construction. Michigan Department ofTransportation.

Muga, H., Mukherjee, A., Milhelcic, J.R., Kueber, M.J., 2009. An integrated assessmentof continuously reinforced and jointed plain concrete pavements. Journal ofEngineering Design and Technology 7 (1), 81e98.

Mukherjee, A., 2011. Project Emission Estimator. The website is available at: http://www.construction.mtu.edu:8000/cass_reports/webpage [accessed 05. 30. 12.].

Navidi, W., 2006. Statistics for Engineers and Scientists. McGraw-Hill, New York.Ozcan-Deniz, G., Zhu, Y., Ceron, V., 2012. Time, cost, and environmental impact

analysis on construction operation optimization using genetic algorithms.Journal of Management in Engineering 28 (3), 265e272.

Peurifoy, R.L., Oberlender, G.D., 2002. Estimating Construction Costs, fifth ed.McGraw-Hill.

Rojas, E., Mukherjee, A., 2003. Modeling the construction management process tosupport situational simulations. Journal of Computing in Civil Engineering 17(4), 273e280.

Rojas, E., Mukherjee, A., 2005. Interval temporal logic in general-purpose situationalsimulations. Journal of Computing in Civil Engineering 19 (1), 83e93.

Rojas, E., Mukherjee, A., 2006a. A multi-agent framework for general purposesituational simulations in the construction management domain. Journal ofComputing in Civil Engineering 20 (6), 1e12.

Rojas, E., Mukherjee, A., 2006b. Multi-agent framework for general-purpose situ-ational simulations in the construction management domain. Journal ofComputing in Civil Engineering 20 (3), 165e176.

RSMeansCostWorks, 2009. http://www.meanscostworks.com(accessedApril, 2011).Santero, N., Loijos, A., Akbarian, M., Ochsendorf, J., 2011. Methods, Impacts, and

Opportunities in the Concrete Pavement Life Cycle. Concrete Sustainability Hub,Massachusetts Institute of Technology, Cambridge MA 02139.

P. Tang et al. / Journal of Cleaner Production 54 (2013) 78e8888

SCAQMD, 2009. Roadway Construction Emissions Model. Available via. http://www.airquality.org/ceqa/index.shtml [accessed 07.30. 11.].

Tang, P., Mukherjee, A., Onder, N., 2010a. Strategy optimization and generation forconstructionproject management using an interactive simulation. In: Proceedingsof 2010Winter Simulation Conference, pp. 3088e3099. Baltimore, MD, USA.

Tang, P., Mukherjee, A., Onder, N., 2010b. Studying dynamic decision-making inconstruction management using adaptive interactive simulations. In: Pro-ceedings of the 2010 Construction Research Congress: Innovation for ReshapingConstruction Practice, pp. 379e438.

Zapata, P., Gambatese, J., 2005. Energy consumption of asphalt and reinforcedconcrete pavement materials and construction. Journal of Infrastructure

Systems 11 (1), 9e20. Special Issue: Sustainability of Transportation and OtherInfrastructure Systems.

Zavadskas, E.K., Turskis, Z., Tamosaitiene, J., 2010. Risk assessment of con-struction projects. Journal of Civil Engineering and Management 16 (1),33e46.

Zhou, M., Pan, Y., Chen, Z., Yang, W., Li, B., 2012. Selection and evaluation of greenproduction strategies: analytic and simulation models. Journal of Cleaner Pro-duction 26 (0), 9e17.

Zou, P., Zhang, Guomin, Wang, Jiayuan, Aug. 2007. Understanding the key risks inconstruction projects in China. International Journal of Project Management 25,601e614.