Decision support for improved construction traffic management and planning

Densifying cities continuously call for new construction, renovation and demolition projects, each generating vast amounts of heavy goods vehicle (HGV) transports. However, how construction transportation affects the urban traffic network remains largely unexplored. This study addresses spatiotemporal network impacts from construction transport, by leveraging traditional traffic and transport simulation. To this end, a framework is presented including (i) a simulation model to compute traffic effects caused by varying off-site construction site transport demands, and (ii) conceptual applications of the simulation model showing construction logistic planning strategies to mitigate congestion disturbances. Simulations are conducted in MATSim using detailed secondary datasets describing site-specific transport arrivals from a case of six projects in Norrk ¨ oping, Sweden. Subsequently, increasing transport demands were assessed on various time-window arrival scenarios against the baseline schedule, which can be used as decision support in urban planning. Results highlight how rigorous construction transport planning avoiding peak-traffic hours can significantly alleviate traffic congestion. This study also emphasizes the need to combine all simultaneous construction projects ’ demands when evaluating disturbances on city-level, alongside the impact on individual links and microenvironments. This study adds knowledge by visualizing the traffic impact during urban transformation.


Introduction
Construction plays a vital role in rendering cities more economically viable and sustainable in the long run.Due to their large and ephemeral nature and immobile location (Ekeskär & Rudberg, 2016), construction sites genereate a vast number of transport trips (P.Josephson & Lindén, 2013;P.-E. Josephson & Saukkoriipi, 2007).In European cities, construction transport represents between 20 and 35% of urban freight traffic (Brussels Mobility, 2016;Transport for London, 2017) and a similar share of transported tonnages in a city (Dablanc, 2009;TNO, 2018).During the construction works, transportation to and from each of these sites impacts the surrounding community (Fredriksson, Nolz, et al., 2021;Ghanem et al., 2018).Construction projects are often treated as a time-limited disruption, but cities are constantly evolving (United Nations, 2018), and as a result, new projects are commenced continuously.Hence, urban areas always have several simultaneously ongoing construction projects; it is only the location that shifts as they commence and finish.The more simultaneous ongoing projects within a city, the more intense their inflicted disturbances are, coming in the form of, for example air pollution, climate change, congestion costs, accidents, noise and economic losses (N.Brusselaers, Huang, et al., 2023a;Guerlain et al., 2019).
While urban freight models often consider transported volumes for a variety of sectors, the link between urban planning and construction volumes, and their effect on traffic flows, is often missing.This paper bridges two current research gaps.First, recent studies have examined the impacts of construction transport at a city level, such as the dynamic health impact of air pollution from construction transport (N.Brusselaers, 2023c), the health impact mitigation potential of construction transport rerouting (N.Brusselaers, Macharis, et al., 2023a), or external costs of the sector's traffic flows at a city level (N.Brusselaers, Huang, et al., 2023a).However, the construction transport's congestion disturbances remain largely unexplored (N.Brusselaers, 2023;Fredriksson, Sezer, et al., 2022).Second, several studies exist on transport modeling and simulation; however, these are often based on aggregated freight demand models (Comi et al., 2012) or do not consider construction transport as a specific transport flow (Sakai et al., 2020;Schröder & Liedtke, 2017).Though, construction transport is temporary in nature, as construction projects need to be supplied during a limited time interval.Thus, their transport flows are unique compared to other urban transport flows.Given these flows do not follow the patterns of other non-temporary flows, construction transportation needs to be treated as a separate transport flow.The temporariness in their routing also makes it necessary to study the impact at both link and city level, as the effect on link level can be substantial near a development area (with several simultaneously ongoing projects).Furthermore, because of the construction transport's daily delivery pattern variations, it is important to capture the sector's specific temporal variations.Conclusively, existing models need to be updated to model construction transport with both better spatial as well as temporal resolution (Sezer & Fredriksson, 2021).Presently, there is a lack of planning tools providing an overview of the total transport demand and the spatial impact of construction transport from a city perspective (N.Brusselaers, 2023;Fredriksson, Sezer, et al., 2022).
This research is positioned on the crossroads between the fields of urban planning, traffic planning and construction transport planning, and aims to address the knowledge gap regarding the spatiotemporal network impacts from construction transport by leveraging traditional traffic modeling and transport simulation (Horni et al., 2016a;Yannis et al., 2006).This study presents a first step in predictively modeling congestion and traffic impacts of future construction sites by introducing a framework including (i) a simulation model to compute traffic effects caused by off-site construction transport to better understand disturbances from construction transportation by indicating distinct effects due to the increased HGV demand, and (ii) conceptual applications of the simulation model to show how construction logistic planning strategies can be elaborated to mitigate their congestion disturbances.The framework is based on a MATSim traffic model utilizing the city of Norrköping in Sweden as a use case for various construction site transport demands and time-window arrival scenarios.
This paper is structured as follows: first, background literature is presented, connecting the identified research gaps in the fields of construction transport planning, urban traffic planning and transport simulation.Based on these identified gaps, a conceptual framework is presented which structures the methodological approach and materials of this study.Finally, results of the simulations and scenarios are presented and discussed, including the study's limitations and future research avenues.

Background
The background section is divided into two main parts, first focusing on the knowledge gap regarding construction transport planning and urban traffic planning, and secondly on the existing traffic modeling shortcomings.

The link between construction transport and urban traffic planning
It is hard to separate construction transport planning from urban traffic planning due to the share of urban transport that the construction transport makes up: 26.40% of HGV traffic in the city of Brussels (N.Brusselaers, Huang, et al., 2023a), and one third of all HGVs in Sweden (Sveriges Byggindustrier, 2010).Construction-related transport must share the infrastructure with other road users (Dablanc, 2007), and the additional transport load from construction increases congestion in the urban infrastructure (Behrends et al., 2008;N. Brusselaers, Huang, et al., 2023a) at the same time as accessibility decreases due to road closures (A.Fredriksson, Eriksson, et al., 2022).Still, construction transport is normally not considered separately in studies of transport planning: see, for example, Gonzalez-Feliu et al. (2012).

Construction transport planning
Construction projects are usually divided into two types: housebuilding and infrastructure projects.This paper focuses on urban housebuilding projects.In such projects, construction logistics can be divided into two primary functions: the management of logistics activities on construction sites (or on-site logistics) and the transport of resources and materials to and from construction sites (or off-site logistics) (Janné & Fredriksson, 2019).The focus in this paper is on the off-site logistics.Transportation is the actual physical movement of goods within the material flow, taking place between two actors, in this case either from the construction supplier/merchant to the construction site, or from the site to a waste management company or mass-storage area.By definition, a construction transport event means two trips: one on the way into a construction site and one on the way out (Fredriksson, Sezer, et al., 2022).A construction site receives 2-10 deliveries, or 8-10 tons of material, per day (Guerlain et al., 2019).Construction transport mostly relies on heavy goods vehicles (HGV) with a maximal authorized mass exceeding 3.5t (N.Brusselaers et al., 2022;Dablanc, 2007;Guerlain et al., 2019).Furthermore, S. mention that personnel traveling to and from construction sites also generate traffic and the use of parking spaces.
The responsibility for planning and coordinating the supply chain and construction site, as well as for keeping disturbances low in the vicinity of the site, resides with the main contractor (Azambuja & O'Brien, 2008).Presently though, the main contractors do not focus on planning transportation (Eriksson, 2019), but rather on regulations, institutional barriers and drivers from within their own organization (Hemel, 2023).Instead, transport planning is often organized by suppliers, as most goods are purchased including deliveries (Bankvall et al., 2010).This leads to a situation of low transport efficiency in terms of fill rates, characterized by empty runs and/or oversized vehicles (Lundesjö, 2015).A common problem for construction transport is time lost in traffic jams (Guerlain et al., 2019;Hyari et al., 2015), as most material deliveries coincide with peak urban traffic (Guerlain et al., 2019).Indeed, non-optimized delivery traffic generates waiting times for unloading procedures and increased congestion levels in surrounding areas (Balm & Ploos van Amstel, 2017;Naz, 2022).The site's location, layout and number of gates play a major role in the transport planning and impact on surrounding transport activities (Guerlain et al., 2019), and an important issue is limited storage space.To limit material accumulation on site, the construction industry has therefore implemented the just-in-time (JIT) delivery system over the last few decades (Lundesjö, 2015).To successfully implement JIT, it has been increasingly common to implement time slot scheduling systems at the construction sites, where time windows are booked for deliveries, enabling coordination of available resources for loading and unloading as well as decreasing traffic jams at, and in the vicinity of, the site.

Urban traffic planning
There exist multiple solutions on how to organize urban freight flows, and especially last-mile deliveries (A.Fredriksson, Janné, et al., 2021;Fried et al., 2023;Sanchez-Diaz & Browne, 2018).These studies have mainly focused on parcel deliveries in combination with emerging technologies, or solutions such as autonomous vehicles, crowd-shipping, parcel lockers, and behavior analysis of e-customers (e.g.Kiba-Janiak et al., 2021;Montwiłł et al., 2021;Pietrzak & Pietrzak, 2021).While inspiration can be gained from these studies, construction transportation is also bound to specific characteristics such as their heavy weight (often excluding alternatives such as bicycles, drones and parcel lockers) and temporary nature in routing.Therefore, so far, construction transport has been treated separately, and logistic scenarios or solutions have been developed (Janné, 2018), such as better planning of supply flows (Thunberg & Persson, 2013), checkpoints with dedicated arrival times (Ekeskär & Rudberg, 2016) or construction consolidation centers (CCCs) (Janné & Fredriksson, 2019;Lundesjö, 2015;Transport for London, 2013), as well as the use of off-peak deliveries (Guerlain et al., 2019;Hyari et al., 2015;Sezer & Fredriksson, 2021).
During construction, a conflict situation arises regarding the city's transport infrastructure, as the transport demand is increasing at the same time as the capacity becomes limited.This conflict situation arises because of a lack of coordination within and between projects and within and between actors, leading to congestion, affecting the living standards in residential areas and decreasing traffic safety for pedestrians and cyclists (Ploos van Amstel & Balm, 2017).Urban planners have a responsibility in coordinating logistics stakeholders in an urban development (Goodman & Hastak, 2006).The decisions made by the municipalities regarding, for example, development areas and zoning plans, restrict decisions made by main contractors regarding layout of sites and locations of gates.However, the congestion impacts of construction transport on the urban mobility network have so far been under-researched, and it is suggested to quantify trade-offs between planning and traffic impacts around the construction site (N.Brusselaers et al., 2022;Hyari et al., 2015).Congestion costs are a negative effect regarding the uncoordinated transportation going to and from the different construction sites, putting an extra load on the traffic network.The main issue posing a hurdle in decreasing disturbances and environmental impact is the lack of understanding on how to plan and control construction logistics within urban transport and mobility planning (A.Fredriksson, Janné, et al., 2021;Lindholm & Behrends, 2012).There is thus a need to see the relationship between the planning of a construction project and its logistics and their inflicted disturbances (Ning et al., 2019), where, for example, Brusselaers et al. (2023b) a) show that the air pollution (NO x , PM) health impact can significantly be mitigated by rerouting existing construction transport flows around spatiotemporal population hotspots.In this regard, impact calculations from Brusselaers et al. (2023a) and Zaalouk et al. (2023) suggest further scenario evaluations.

Traffic modeling and urban freight models
Traffic modeling has previously been extensively applied in the context of city logistics and urban freight (Yannis et al., 2006).In this regard, multiple macroscopic urban freight models have been built, which are, among others, based on surveys carried out as part of national programs (e.g.FRETurb V3, (Routhier & Toilier, 2007)), statistical shipments data (e.g.CBS), aggregated (regional) data on urban freight (e.g.Eurostat), commercial services data, or a set of socio-economic drivers (population, Gross Domestic Product, trade, etc.) (e.g.ITF urban freight transport model, PASTA (ITF-OECD, 2023).These models aim to assess urban freight movements given different policy measures and their evolution over given time frames, with output on transport flows, vehicle characteristics and other transport-related (sustainability) indicators.Their use is often intended for urban planners to estimate the road occupancy (parking and traffic) by freight vehicles, and to simulate the evolution of this occupancy according to assumptions regarding defined facilities within the agglomeration.Also, national-or regional-specific freight models have been developed, such as the TRansport Agent-BAsed Model (TRABAM) (Mommens et al., 2018) for the territory of Belgium, which is based on the open-source software MATSim (W Axhausen et al., 2016).It observes the equilibrium resulting from simulating the interactions between the various goods transport operators in a multimodal network and can be linked to external cost calculations or the impact computation of, for example, off-hour deliveries (Mommens et al., 2018).
Through overviewing these existing models, it is found that construction transport is either (i) not rendered in such models at all, mostly because of the lack of knowledge on the transport-attraction for construction site projects and the varying spatiotemporality of construction sites, or (ii) are included as part of a holistic vehicle fleet, but cannot be dissociated from other transport sectors on a fine granular level other than in their role as an economic commodity (e.g.NSTR subclasses) (e.g.third-party logistics providers transporting non-exclusive constructionrelated goods).N. ) have, though, developed algorithms to filter construction-specific vehicles from total HGV traffic at a city level, and historically calculated the sites' monthly transport-attractions, and the sector's share and transport impact in terms of external costs.However, a forecast of construction site transport demand, and an investigation of which bottlenecks arise depending on timing and location of sites, have not been attempted so far.
This gap can nevertheless be approached by combining the fields of urban traffic planning with construction transport planning, and by leveraging traditional traffic modeling.Fig. 1 presents a conceptual framework based on the identified research gaps, structuring the methodological steps employed in this study (indicated in green).From an urban planning perspective, the modeling of construction transport should contextualize spatiotemporal relations between aggregated flow, density and speed (Burghout et al., 2006).Given the scope of this study, a mesoscopic model is necessary to grasp the properties of both microscopic, such as individual vehicle characteristics (Khan & Gulliver, 2020), and more macroscopic models.These are based on probability distributions and queuing theory for the modeling of traffic flows, enabling the description of traffic-flow behavior in aggregated states (Khan & Gulliver, 2020).Furthermore, multiple ways exist to define road congestion (Grant-Muller & Laird, 2007).Typically, congestion is described as the impedance that a vehicle imposes on the (users of a) network, when coming closer to the maximum network capacity (Goodwin, 2004).One option to measure congestion levels is to calculate travel delays for road segments.Such an approach has been applied in a study by C. combining GPS data with MATSim to estimate passenger vehicle delays calculated as the difference between actual travel time and free-flow travel time.In this study, link delay was chosen to be the main measure to describe the congestion effects, since a link-level view is required to accurately assess external effects.Travel times can be estimated using GPS probe data (Zheng & Van Zuylen, 2013), Bluetooth sensor data (Yildirimoglu, 2021), or a combination of sensors (L. ) and predicted using data-driven approaches (Jenelius & Koutsopoulos, 2017), model-based approaches (Oh et al., 2018), or a combination of them (Allström et al., 2016).MATSim is a framework used for mesoscopic traffic modeling known for its computational capacity and ability to simulate scenarios on a 24-hour basis (Horni et al., 2016b).In the MATSim model, each agent optimizes its execution of activities throughout the day as iterations increase, while competing for space-time slots in the traffic network.This optimization of plans is done through different choice dimensions, such as planning of route, time and transport mode, leading to a stochastic user equilibrium based on coevolutionary principles (Horni et al., 2016b).This is then linked to application scenarios to evaluate and compare the performance.

Methodological framework and materials
The presented framework in Fig. 1 bridges the two identified research gaps: how (future) construction transport flows affect the urban traffic network and its concomitant congestion effects, and the lack of spatial planning tools due to the lack of inclusion of construction site transport demand in current simulation models from a city perspective.The present study follows a sequential exploratory mixed methods design (Pluye & Hong, 2014), divided into three main steps.First (cfr.Section 3.1), the use case of the city of Norrköping, Sweden is explained, and the forecasting of construction site transport demand is calculated.Second (cfr.Section 3.2), construction site information is combined with traditional and validated traffic modeling using MATSim (Horni et al., 2016a;Yannis et al., 2006) to evaluate congestion effects.We calculate multiple transport demand scenarios, which, third (cfr.Section 3.3), form the basis for their spatiotemporal distribution.The construction transport simulation output provides network and construction traffic effects, and generated truck traffic and infrastructure effects.These computed construction-related flows are subsequently interpolated with the baseline mobility traffic to identify network bottlenecks, traffic effects and link travel times.Next, the interplay between increasing transport demand scenarios and time-window arrival scenarios for the construction transportation is computed.Results are expressed as total N. Brusselaers et al. daily delay, total delay divided in one-hour time bins, and average link delay.These were based on the total number of transport flows traveling on a link and expressed as a percentage compared to link free-flow travel time, or as total delay for hourly time bins, or for one day.Since this paper's purpose is to measure the effect of HGV flows on congestion, the focus is on links with HGV traffic.All results are evaluated by gradually adding sites in the traffic model.Such an approach allows us to understand the effects of different construction sites with varying estimated HGV flows.

Use case and forecasting transport demand
Step 1 in Fig. 1 is forecasting the transport demand.The aim of using Case Norrköping is to prepare datasets that describe transport flows for HGV, resulting in datasets from which to create a method that allows modeling of disturbances.Norrköping is a mid-size Swedish city with N. Brusselaers et al. several large development projects planned, such as a new area close to the old port, a new railway station, and a new high-speed railway connection to Stockholm.This context is therefore suitable as a singlecase study (Ying & Tookey, 2017).Norrköping has already several ongoing construction projects for residents in different areas, such as Kvarntorget and Ektorp (Norrköping, 2020).Going hand in hand with anticipated population growth, new residential areas such as Inre Hamnen and Himmelstalund will be developed.It is therefore important to analyze how to handle the arising challenges in the urban environment caused by construction projects.The use case of Norrköping estimates traffic disturbances from six forthcoming construction projects in Norrköping that are active in parallel over a considered 51-month period, as represented in Fig. 2. Two are in the southern part of Norrköping, close to the intersection of highways E22 and E4 -Söderporten 1 and 2, and four are in central Norrköping -Inre Hamnen, Hotell Svea, Tingsrätten, and Spinnhuset.
Construction site transport demand needs to be predicted for these simultaneously active urban construction sites.This is based on historical data patterns (e.g.construction site variables such as gross floor area, type, time plan, sensors or raw transport data) or forecasted data input (such as the estimated traffic flow of different projects, time frames or terminal data).Forecasting, a well-known technique employed in various fields, helps predict behaviors, patterns, trends and fluctuations (Petropoulos et al., 2022).
In this study, secondary transport data gathered and documented by A.A. from twelve construction projects in Sweden between January 2017 and April 2020 were used to feed the model.The data was obtained from gates and booking calendars and contains information about the arrival time of transports at construction sites and common transport dimensions.Table 1 summarizes the data variables used in this study and in the MATSim simulation, with Simulation defining if the variable was included in the simulation study, Data clarification describing how the variable was used, and Source describing which data collection method was used to acquire values.The MATSim input population was generated using an Origin-Destination (OD-)matrix based on mobile network data describing traffic flows in Norrköping as depicted in N.
When creating the HGV OD-pairs, first, the total estimated transport demand was converted to transport arrivals for one working day.Estimated demand was expressed as the number of transport trips per month for the whole construction period and was based on historical transport data for seven construction sites in Stockholm, for which a coefficient was calculated based on the site's gross floor area.Given the gross floor area and construction period for construction sites in Norrköping, the demand for the six considered sites was calculated to vary between and 3845 HGV trips per site.The destination points were set at the construction site locations.The total transport flows were based on estimated transport arrivals at each respective site, and their return flows based on their origin locations.
The total estimated transport demand across the six considered construction sites in Norrköping equals 15,896 HGV transport trips.On analyzing the secondary data, it was decided to create worst-case scenarios based on the day with most HGV arrivals: hence, the pessimistic modeling of transport demand scenarios.Three gradually increasing daily transport demand scenarios were calculated, based on the transport arrivals as a percentage of total site demand.The first (0.96%) corresponds to the daily transport demand from most construction sites.The second (2.8%) is based on the average share for all sites.Finally, the third scenario (8.8%) is based on the day with most transport arrivals of all sites.This leads to estimated daily absolute transport demand scenarios of 152 HGV, 458 HGV and 1404 HGV transports.While the likelihood that all sites have their worst-case day simultaneously is low, the two last scenarios are used to indicate distinct effects that would otherwise be overlooked with smaller HGV demand.
The origin point for construction transport trips was based on assumptions regarding estimated demand from local-and long-haul suppliers, of which the distribution is based on traffic volumes from Traffic Web (2023).The data allowed to determine the time periods when transport arrivals are scheduled.The hourly distribution of transport arrivals at the construction site is presented as baseline HGV time-window arrivals (LS0).Due to trade secrecy, it was not possible to create OD-pairs for HGVs between suppliers and destination based on actual data.Instead, it was assumed that every construction site can be served by any supplier.For that reason, an algorithm based on secondary data was developed that creates OD-pairs, which can be found as part of.Trips associated with construction sites were divided into either long-haul or local.An assumption was made that long-haul transportation with an origin outside the border of the municipality of Norrköping represents half of all demand, and origin locations were set to exits of incoming highways E4 and E22.The remaining half of estimated transport flows were assumed to be local, from three main construction material warehouses in Norrköping -Optimera, Renall and Beijer Byggmaterial, with origins within Norrköping's boundaries.The origin and destination points were visualized in Fig. 2. Data on personnel transport was also available, sourced from a travel survey in Linköping (A.A. Sezer & Fredriksson, 2020).However, simulation runs based on the number of workers needed on site (with a total of workers for the six considered sites), their spatiotemporal travel patterns and approximate home locations and transport mode (72% by car), highlighted that the personnel trips only account for a 0.2% increase compared to baseline.Due to this negligible impact, it was determined to exclude workers from the simulation runs in further analyses.

Construction transport and urban traffic modeling
Step 2 in Fig. 1 is the computed construction site transport demand combined with traditional traffic modeling, using MATSim (2023) to assess the congestion effects caused by construction transportation, in combination with QGIS 3 (QGIS, 2023) for verification purposes.The process followed in the MATSim model is visualized in Fig. 3. Initial demand consists of agents representing the studied area and their daily activity schedule.This step also involves preparation of other input files, such as a traffic network of the study area and a configuration file of the simulation.The configuration file contains specifics about all involved steps in the MATSim loop.mobsim is responsible for network loading.scoring evaluates the executed plans, and the replanning step tries to optimize the chosen plans.When the simulation is finished, different analyses can be applied to the produced output.The length of simulation depends on the number of iterations and should be continued until the average score calculated in scoring has stabilized (Horni et al., 2016b).
First, it was an iterative process for the MATSim model to find a realistic traffic state for cars in Norrköping.To create the population in Norrköping, an OD-matrix estimated from mobile network data was used and converted to the input needs of MATSim, based on a method provided by N.. Trips were created by extracting mobile network data describing people's movements, consisting of flows between 189 traffic analysis zones aggregated at an hourly level.MATSim produced output plans describing car flows, which were compared to link flows in Norrköping acquired from Traffic Web (2023) as the number of vehicles per hour at given locations.shows the flowchart of the processes involved in the MATSim model combined with trips from Case Norrköping.To run the MATSim simulation, three files were compiled.First, a network file describing the traffic infrastructure was built, including link length, capacity, speed limit, permlines and transport modes, which were based on data provided by OpenStreetMap.The network was updated with construction sites and local suppliers.
Second, the population file described the travel demand, with agents and their daily plans, including activity type, location, defined ending time and transport mode.The transport mode is kept until (if at all) another transport mode is chosen in the replanning stage.The activity ending time defines when an agent starts moving to the next activity (Horni et al., 2016b).Routes were calculated based on the shortest path between two activities (Rieser, 2019).The complete population file was created from the OD-matrices describing the travel patterns for people in Norrköping and the HGV going to and from construction sites.Daily activity chains were created for every agent.People in Norrköping were described using a simple activity chain consisting of two activitiesstart and end, indicating starting and ending time for one trip.HGVs receive an activity chain consisting of three activities consisting of departure time from home or suppliers, time spent at construction sites, and departure time from construction site.The information provided was filtered for one working day.
HGVs going to and from a construction site were modeled using their respective OD-matrix, arrival and departure times, and transport mode.Two HGV types were simulated (A.A. Sezer & Fredriksson, 2020): approximately 80% of transport arrivals are HGV with a length of ≥10 m, and 20% with a length between 19 and 24 m.The speed was capped at 80% of link free-flow speed for HGVs ≥10 m and at 65% for larger HGVs.Each HGV agent contains three activities: (a) the process of leaving its origin toward the construction site, (b) the process of unloading (or turnaround time), and (c) the process between the construction site and its starting location (i.e.going back).The length of turnaround time depends on the supplier type and varies between 30 and 60 min.Long-haul was assumed to take 60 min, and transport trips from local suppliers have a shorter turnaround time identified in the SUCCESS project as well as historical data from previous projects (Guerlain et al., 2019;Sezer & Fredriksson, 2021).Lastly, a configuration file was built, defining how the simulation was configured.
Third, Mobsim was used to load the network and estimate the traffic flow.Traffic dynamics were modeled with waiting queues that were updated in each time-step (Horni et al., 2016b) based on the first-in-first-out (FIFO) principle.This principle holds that fast vehicles cannot pass slower vehicles.The advantage of such a queuing approach is the computational efficiency, which is useful for large-scale scenarios.A disadvantage is that this approach may not fully consider car-following effects (Horni et al., 2016b).Therefore, a handling system for very congested conditions, when agents are barely moving or not moving at all, was used.After a pre-defined interval of no movement, an agent is forced to enter the next link while ignoring the storage capacity constraint.Such a solution avoids gridlocka state in traffic congestion where no vehicle can move forward (Rieser, 2019).All executed plans were valued using a score calculation.Scoring is a central element of MATSim, since the choice of each selected plan is based on a score.Typically, more iterations give a better optimized daily schedule plan for all agents, and a state of equilibrium is reached when the agents are not able to further improve their daily plans (Horni et al., 2016b).This replanning phase should work in a way that undesired plans are removed, and plans with a high score are kept (Rieser, 2019;Tchervenkov et al., 2020).Every strategy was coupled with a weight that defines the probability of applying a certain strategy to the incoming set of agent plans.Therefore, it was important that scoring calculations represent the goals of the simulation and studied area.This was handled by the scoring function formulated by D. presented in Eq. ( 1). (1) The scoring function provided for a plan is based on the sum of activity utilities ∑ N− 1 q=0 S act,q and travel utilities ∑ N− 1 q=0 S trav,mode(q) (Horni et al., 2016a).The activity utility is then calculated as follows: S act,q = S dur,q + S wait,q + S late.ar,q+ S early.dp,q+ S short,dur,q (2) Where S dur,q is the utility of performing an activity, S wait,q the utility of extra waiting time, S early.dp,q the utility of leaving an activity earlier, S short,dur,q a penalty for an activity ending too early, S late.ar,q the utility for late arrival (Horni et al., 2016b).S dur,q is calculated following the expression: S dur,q = β perf * t typ,q * ln ( t dur,q t 0,q Where β perf is the marginal utility of activity duration, t typ,q is the typical activity duration, t dur,q is duration of performed activity and t 0,q corresponds to the duration until utility gets positive.The marginal utility factors in S wait,q , S early.dp,q and S short,dur,q were set to zero. Travel utility S trav,mode(q) for a trip q is calculated by the expression: S trav, mode(q) = C mode(q) + β trav,mode(q) * t trav,q + β m * Δm q + ( β d,mode(q) + β m * γ d,mode(q) ) * d trav,q + β transfer * x transfer,q (4) Where C mode(q) is a mode-specific constant, β trav,mode(q) the marginal utility of traveling time depending on the mode, t trav,q the traveling time between two activities, β m the marginal utility of money, Δm q the change of budget (e.g.road tolls is an example of β m * Δm q ), β d,mode(q) the marginal utility of distance, γ d,mode(q) the monetary distance rate specified by mode, d trav,q the traveled distance between two locations, β transfer the penalty for public transport, and x transfer,q the decision variable associated with transfer penalty (Horni et al., 2016b).N. Brusselaers et al.The model was calibrated and validated based on 80 iterations using the visualizing tool VIA (Simunto, 2023) andTraffic Web (2023) link counts for 61 locations in Norrköping, and recorded using a pneumatic tube over a road segment (Traffic Web, 2023).The comparison was made for a full working day.Such a validation approach is common according to J. The overall flow profile for cars was consistent and follows measurements with a 3% difference between the simulated passes (556,000) and the measured observed passes (574,000).It must be noted that there exists variety and some locations in MATSim has the risk for increased congestion due to car volumes that are too large, for which their situations were adjusted.Further, a statistical analysis of all measurements at every hour gathered from 61 counting station locations was conducted.The analysis consisted of a regression plot to determine R 2 value and mean absolute percentage error (MAPE) for every measurement expressed as a histogram.A regression plot, available as part of was made for all measurements at every hour.It compares link counts per hour in MATSim with statistic counts for car flows.R 2 value was around 0.78, indicating that the model explains a reasonable amount of variance of observed flow volumes in measurements.However, further calculation of MAPE identified the larger measurement errors to occur at night-time (larger differences on smaller values).Such errors during night-time are acceptable and do not affect the results of this study.

Network effects and scenario evaluations
Step 3 in Fig. 1 shows that scenarios of arrival time window scheduling were implemented to test the effects of controlling construction transport using various HGV routes.From here, these are referred to as HGV time-window arrival scenarios.Three scenarios were applied, as shown in Fig. 4. The scenarios were compared against a car delay baseline, combined with the actual arrival rate of materials to site on an hourly basis for that day (LS0), and thereafter compared to each other by computing total delay for 24 h on HGV routes and expressing total delay in hourly time bins.
• The first arrival time window scenario (LS1) aimed toward a constant arrival rate at the construction sites throughout the day between 5:00 and 21:00.• The second arrival time window scenario (LS2) represented a distribution scheme with avoidance of peak hours, i.e. 6:30-9:00 and 15:00-18:00.The arrival period for early morning is set between 4:30 and 6:30.• The third arrival time window scenario (LS3) strictly follows the concepts of peak-hour avoidance, in combination with early morning or late evening deliveries.

Results
To assess the effects of construction transportation on urban traffic, first, spatiotemporal analyses were conducted that considered increasing daily transport demand scenarios (152 HGV, 458 HGV and 1404 HGV).Subsequently, the HGV capacity and supply of the traffic network were calculated.Next, alternative HGV time-window arrival scenarios (LS1, LS2 and LS3) were assessed.Finally, sensitivity analyses were conducted to understand the influence of random values on the produced results.The results involve the analysis of six construction sites with corresponding estimated HGV transport demand and are aggregated as total congestion hours expressed in one hourly time bins.To understand the significance of the modeled congestion effects, all situations are compared to a baseline, expressed as the total congestion level for cars without the addition of construction transport, as shown in Fig. 4. It shows the present baseline delay hours throughout one day, approximately 1314 h, of which the HGV-allowed routes total 284 h.These delay volumes are based on approximately 313,000 car trips executed in the simulation, averaging 15 s delay per trip.Fig. 4 also presents the baseline distribution of arrivals of HGVs at construction sites per hour over the day (LS0).

Temporal analysis of transport demand
Fig. 5 shows the congestion levels on both the whole traffic network and on selected HGV routes, for the various transport demand scenarios 152 HGV (A.1 and A.2), 458 HGV (B.1 and B.2), 1404 HGV (C.1 and C.2). Cars represents the baseline level without construction transportation.The congestion effect when all sites are included (site numbers 2-7) is represented by Sites: 743,562 HGV.The latter is divided into Sites: 7435 to describe the Northern area of the city, and Sites: 62 the Southern area.In general, it is concluded that a larger number of HGVs have a greater impact on car congestion levels, an expected result.Because the congestion impact is rather minor when assessing the complete traffic network (Fig. 5, transport demand scenario A1-C1), it was decided to also separately analyze selected HGV-allowed routes (Fig. 5, transport demand scenario A2-C2), which is also the focus of subsequent analyses in this paper.
• In transport demand scenario A.1, the increased delay of 1410.7 h (or 7.4%) is considered minor with the introduction of construction transport.A slight increase can be noticed after 13:00, when all sites are being served.Focusing solely on the HGV routes within the network, scenario A.2 brings more detail compared to the complete traffic network, where total delay hours increase from 248 h to 371 h compared to the baseline.• In transport demand scenario B.1, a small increase can be noticed at morning peak hour (8:00) and continuing until 13:00.After 13:00 the delay starts to decline around the baseline level.Across the whole system, the congestion effect is not significant, but still increases delay hours by 23.5% to 1623 h.Scenario B.2 clearly demonstrates congestion effects, with a 128.2% increase compared to baseline, especially during morning peak hours.• In transport demand scenario C.1, effects on whole traffic system level can be distinguished.An increase in delay hours is clearly shown in the morning peak hour at 8:00.A distinct impact on congestion with an increase of 78.9% compared to baseline is shown, totaling 2351.26 delay hours.In scenario C.2, total delay hours increase by factor 4, totaling 1263 h, with noticeable increased congestion in the morning peak hour at 8:00.Also, a pattern arises, with an increase in congestion hours in the afternoon.

Spatial analysis of transport demand
Fig. 6, scenarios A1-C1, show absolute HGV link volumes when all six sites are served.Congestion effects for HGV-allowed routes are mapped for the various transport demand scenarios 152 HGV (A), 458 HGV (B) and 1404 HGV (C), and scenarios A2-C2 show the average HGV link delay, expressed as a percentage against the free-flow link travel time.
• In transport demand scenario A, average delay increases are limited.
The Southern area attracts a large share of construction transport demand, which translates into higher HGV intensity on incoming routes to these sites (>70 HGV).Sites in the Northern area is more scattered, resulting in lower HGV intensities in the site's immediate vicinity.• Transport demand scenario B shows that a couple of construction sites with a low transport demand (88 HGV) do not contribute to significant delay increase (<5%).When all sites are gradually added, noticeable effects start to emerge, with several short links experiencing a significant delay increase (5-10%).• Transport demand scenario C results in very high HGV volumes, with most links experiencing >100 HGV passes throughout a working day and a delay increase of >10%.Several roundabouts and connecting links have a distinct delay increase between 10 and 30%.
Fig. 6 indicates that link delay increase is not directly proportional to HGV volume on links, and that the congestion increase is varying and depends on the exact link location.Furthermore, one road, Ståthögavägen, is shown to be very sensitive to HGV intensity and is therefore expected to be a bottleneck when adding new construction projects into the urban planning of Norrköping.

Network supply capacity and gridlock limit
Based on the results above, the maximal HGV supply point of the network can be determined for (i) a HGV demand level at which no significant congestion impact is caused, and (ii) a level at which HGV demand causes traffic to move toward gridlock.Findings show that the delay increase is almost linear with the increase in the number of HGV below 1000 HGV in the system.Fig. 7 shows the total delay hours divided over hourly time bins with an applied step size of 500 HGV.The major tilting point can be set at 2000 HGV, where the increase of delay gets significantly larger and hits a peak of undesirability.The traffic system state approaches a breakdown when >3500 HGV are introduced into the system, causing exponential increases of total delay hours.

Application of HGV time-window arrival scenarios
To test the effects on the traffic system of altering transport arrivals, HGV time-window arrival (or arrival-time scheduling) scenarios were applied, as presented in Fig. 4. LS1 follows a constant arrival rate between 5:00 and 21:00; LS2 avoids deliveries during peak hours between 6:30 and 9:00 and between 15:00 and 18:00, and LS3 strictly follows avoidance of peak hours through a combination of early morning and late evening deliveries.All three scenarios are compared against the baseline actual arrival rate of materials LS0, and were applied to increasing daily transport demand scenarios (152 HGV, 458 HGV and 1404 HGV).Fig. 8 presents the difference of total delay for cars (with baseline 284 h of delay) for the three time-window arrival scenarios, under different HGV demands, i.e. 152 HGV (A), 458 HGV (B) and 1404 HGV (C).After 12:00 similar results to LS1 and LS2 are obtained.These reduce overall delay by respectively 2.03% (367.77h) and 6.56% (350.76 h) compared to LS0, and this is mainly attributable to lower congestion in morning peak hours.However, increased delay levels in afternoon peak hours are noticed.LS3 generates 337.88 h of total delay time, or 9.99% lower delay time compared to LS0.LS3 increases delay levels between 9:00 and 13:00 but matches the delay profile in afternoon peak hours (16:00-19:00).• When transport demand is increased to scenario 458 HGV (B), the variation of total delay for certain hours between arrival scenarios follows scenario 152 HGV (A), albeit in a more pronounced way.At this stage, LS3 is considered the best solution, providing a congestion level decrease of 21.31% (433 h) compared to the baseline LS0 (551 h), followed by LS2 with a decrease of 13.33% (477 h).Despite being a simple solution, LS1 also provides a minor improvement of 2.17% (539 h) compared to LS0. • When transport demand is further increased to scenario HGV 1404 (C), LS0 shows itself to be an inappropriate solution for the morning peak-hour period, and all alternative scenarios provide a better result.As for afternoon/evening, the effect of LS1 and LS2 results in a significant increase in congestion at 16:00.The benefits of LS3 are very distinct given the difference for congestion hours in afternoon peak hours and comes out as the best solution (totaling 721 h of delay, or a 41.64% decrease compared to LS0), followed by LS2 (941 h; − 23.76%).
Conclusively, the results show that the baseline LS0 solution may not be efficient from a congestion perspective compared to implementing some type of time-window arrival scheduling as tested in LS1-LS3.

Sensitivity analysis
Sensitivity analyses were conducted by running all considered scenarios with ten iterations, as presented in Fig. 9.It shows the variation of total delay between the different iterations.Random decisions in the model influence the result at certain hours, but overall, the profile of total delay is consistent.The variation of the results is very minor for scenario HGV 152 (A) and gets slightly more distinct when the number of HGV increases in scenarios 458 HGV (B) and 1404 HGV (C).Fig. 9 shows that results have a higher variation between 8:00 and 9:00 and between 15:00 and 17:00.Table 2 presents how random values affect total congestion hours on 24-hour runs, and highlights that the effect is not large given the greatest difference between minimal and maximal measured total congestion hours is 13.8% in the most extreme scenario.Hence, random values do not affect the general conclusions, comparisons, and interpretation of the presented results.
The next section further interprets and discusses the results in comparison with existing literature, with room for its limitations and further

Discussion
The contributions of this study are (i) a simulation model to compute traffic effects caused by off-site construction transport to better understand disturbances from construction transportation by indicating distinct effects that would otherwise be overlooked with smaller HGV demand, and (ii) conceptual applications of the simulation model to show how construction logistic planning strategies can help to mitigate congestion disturbances.The modeling of urban mobility effects from construction transport is built on existing and established ways of working using the open-source traffic modeling software MATSim (MATSim, 2023).However, the novelty is the focus on construction transport, an earlier excluded urban freight demand, which, as seen in this study, has a potential large impact on congestion depending on the number of simultaneously ongoing sites and their distribution in relation to each other (see Figs. 5 and 6).The paper contributes by presenting a structured way to model construction-related transport trips based on limited data, and therefore increases its applicability as part of strategic transport planning in urban planning processes.The congestion impact of construction transport has previously been a neglected decision variable, as models so far have only supported the regular traffic without the induced increased traffic of HGV on routes to and from construction sites.The few existing prior studies assessed traffic emissions (using life cycle assessment) (Huang et al., 2009) or delay impacts (Lee et al., 2005) due to disruptions caused by road maintenance.We also show the importance of hourly time buckets and a link-level view to identify high spatiotemporal discrepancies when running simultaneous construction sites in the city, which in turn, can be linked to generated (congestion) disturbances in microenvironments (such as schools).This allows for a broad urban overview but also identifies local impacts and how these relate to spatiotemporally overlapping flows related to other freight or mobility sectors, and how changes in the local network supply (e.g.road closure) affect and/or propagate (negative) effects.However, while the model still needs further validation, it gives an indication of the need to also consider construction transport in urban transport modeling.We therefore highlight the need for further research into the interplay between the conducted mesoscopic analyses on a city level (both on the whole network and the HGV routes) and on the importance of  considering vicinal microenvironments (on a link level) from a multi-disturbance perspective, and the consideration of total construction site transport demand (all simultaneous active sites) within a given time frame and geography.Within the research field of off-site construction transport impact assessments, this study focused on congestion and traffic disturbances.In recent years, other externalities such as air pollution (Brusselaers et al., 2023b), climate change (Chelly et al., 2019;Sezer & Fredriksson, 2021), noise (Rönnberg et al., 2022) and accidents have been the source of investigation.These studies show that the fastest route might not always be the most optimal one from a disturbance perspective, as longer routes might decrease overall air pollution exposure from construction transport by avoiding high and/or vulnerable population densities (Brusselaers et al., 2023b).One individual site does not impact congestion levels at a significant level.However, the combination of several sites close to each other does show significant effects, thereby validating the impact zones as presented by Fredriksson et al. (2021a).In this sense, the observations of this paper follow findings of Mommens et al. (2019) and Brusselaers et al. (2023b), who highlight the importance of considering microenvironments (such as schools) in the dynamic environmental assessment of urban freight traffic.The current evaluation of six construction sites in Norrköping was evaluated by means of proof of concept of the traffic modeling framework.This leaves room for future research to consider a combination of these effects to gain a more holistic view on construction transport-related disturbances in a city.
This paper presents the first step in forecasting disturbances, i.e. the impacts on future construction site transport demand, from an urban planning perspective, by demonstrating the feasibility of using the results to show congestion, traffic network and construction transport arrival scheduling effects from construction transport in both time and space.It presents how detailed congestion disturbances can be simulated using limited data.The results emphasize the avoidance of peak hours and suggest more extensive use of early morning and late evening deliveries.This puts additional requirements on the construction industry as the use of, for example, LS3 may be too difficult to adopt.In this regard, results of LS2 may be more realistic in terms of its practical implementation.Further detail could render stronger insight into the actual route and next destination of an HGV when it is arriving at site.A flat distribution could be used as a booking calendar and system with available timeslots.Another advantage associated with LS1 is the steady and predictable incoming flow to construction sites that can be used to increase efficiency at a site.There is also room for future research to examine how predictive models can be implemented to anticipate inflicted disturbances from construction logistics and improve planning efficiency on construction sites to come, in coordination with other mobility and freight network users.In this context, the present implementation of the proposed MATSim model in Norrköping has room for improvement.The complete MATSim loop with replanning stage was only applied to find a state for car traffic flows in Norrköping that represents reality at a reasonable level.With a better calibrated model, the functionality of the replanning stage could be applied to analyze situations from a long-term perspective.Additionally, in the creation of ODpairs, not all arrival combinations were tested, and better combinations may exist.However, OD-pair creation is considered fair for large HGV demand.Similarly, the current study presents data aggregation in hourly time bins, which is common, but more narrow time bins could be envisioned in further studies.
Within the broader field of urban freight traffic, the developed method in this study could be used for any HGV input.This is in line with an ongoing research stream whereby increased knowledge is sought to further investigate the interplay, synergies and synchronization between urban freight sectors and urban mobility traffic (e.g.Kiba-Janiak et al., 2021;Montwiłł et al., 2021;Pietrzak & Pietrzak, 2021) and its related equity and disturbance assessments (e.g.N. Brusselaers, Huang, et al., 2023a;Fredriksson, Sezer, et al., 2022;Fried et al., 2023).This paper covers off-site activities within construction logistics, hence making abstraction of on-site logistic activities, and the dynamics between these two (Naz, 2022).Moreover, a rigorous procurement and on-site storage planning also show positive interdependencies with construction productivity and project profitability (Vu-Hong-Son & Huynh-Chi-Duy, 2023), hence solidifying the importance of considering off-site construction flows with on-site efficiency gains (Naz, 2022) and integrated planning or scheduling systems between off-site and on-site operations (A.Zaalouk et al., 2023).Examples in Sweden indicate that a system to keep track of on-site transport arrivals is often lacking.By adding a construction logistics plan, increased efficiency regarding coordination and planning can be achieved (Andersson & Van Heek, 2023).In doing so, the construction site layout play an important role (Andayesh & Sadeghpour, 2014;Hammad et al., 2016;Sanad et al., 2008), which vice versa have an impact on construction transport arrivals and transport network congestion costs in the immediate vicinity of the site (Song et al., 2018).In this regard, a link can also be made to current developments of interactive urban freight decision support tools.An important aspect is to support collaborative planning of construction transport in urban planning by means of interactive visualization, as presented by Fredriksson et al. (2022a).Hence, it provides a shared deliberation platform to discuss identified alternatives and disturbances among the involved (public) stakeholders.This tool can present more insight into base cases and key disturbance effects of construction logistic flows, incorporating the results of computed traffic scenarios across multiple disturbances (Fredriksson, Sezer, et al., 2022).

Conclusions
The overall aim of this paper was to address the knowledge gap regarding the spatiotemporal network impacts from construction transport, by leveraging traditional traffic modeling and knowledge in the field of transport simulation by developing a framework to assess the congestion effects.This can later be used as decision support in urban planning.The study showed that construction transportation contributes to an increased congestion level.The results show that at a local level (certain parts of road or selected routes), the congestion effect is more significant.The results emphasize the importance of considering two aspects: the total number of construction sites within an area when potential disturbances at municipality level are evaluated, and the study of separate links.Also, construction project size should be considered, as it provides an indication of the arriving volume of HGVs at construction sites affecting certain areas.Conclusively, this study stresses the importance of actively managing construction logistic, and more specifically, the use of time-window arrival planning.The results clearly indicate that planning arrivals to avoid peak-traffic hours can alleviate the urban road network.Even a constant hourly arrival rate at the site shows improvements over the current baseline delivery scheme.The potential benefits of the planning schemes become more distinct with larger HGV demand.The development of the presented planning system adds knowledge by visualizing the traffic impact of development and traffic plans during urban planning.N. Brusselaers et al.

Fig. 6 .
Fig. 6.HGV link volumes and average HGV link delay for transport demand scenario 152 HGV (A), 458 HGV (B) and 1404 HGV (C) when all six sites are served.

Table 2
Sensitivity statistics for total congestion hours on HGV routes in use case.