Cities have been growing at a rapid rate since the past few decades, and so has their need for transporting goods and people. According to the United Nation’s forecast, by the end of 2050, around 68% of the world’s population will be the urban population and the rate of urbanisation will be much faster in developing countries, such as India and China (UN DESA 2019). The movement of goods in any city, that is, urban freight transportation (UFT), plays a vital role in the city’s economy and lifestyle, whether it is a supply of goods or collection of waste. The European Commission defines UFT as the movement of the freight vehicles whose primary purpose is to carry goods into, out of and within urban areas (Transmodal 2012). The majority of the freight vehicle trips in urban areas are because of commercial activities (Holguín-Veras & Jaller 2014). The pressure of population growth in urban areas has resulted in the spread of city boundaries and logistics sprawl, that is, spatial de-concentration of logistics facilities and distribution centres in the metropolitan regions (Dablanc 2011), which are responsible for more and longer freight vehicle trips.

The share of freight vehicles in city traffic is generally low. However, because of different physical and dynamic characteristics (like slow acceleration and sheer size) as well as operational behaviour (like consumer-oriented delivery time, multiple stops, etc.), their contribution towards negative impacts (congestion, air and noise pollution, accidents, climate change and infrastructure damage) is significant (Browne et al., 2007, 2012; Dablanc 2007; Ewbank et al. 2020; Goel & Guttikunda 2015; Ogden 1992; Taniguchi & Thompson 2018; Transmodal 2012). These impacts result from interactions amongst goods, facilities, infrastructure and vehicles in four subsystems - accessibility, land use, transport and traffic integration (Lindholm 2012). Some of the impacts of urban freight are limited to the area where the vehicle movement takes place, and these effects last for a short duration of time, for example, noise pollution, congestion-related problems, infrastructure damage, etc., whilst other impacts are not bounded, and they affect a larger area over a long period like environmental emission and climate change (Behrends 2011; Behrends, Lindholm & Woxenius 2008). The scale of local impact is proportional to the population density in that locality (Maibach et al. 2007). Freight traffic represents around 20% – 30% of vehicle kilometres travel and 16% – 50% (depending on the pollutant considered) of the emission of air pollutants by transport activities in a city (LET-Aria Technologies 2006). In developing countries like India, urban freight vehicle travel of total metropolitan vehicle travel is substantial with 37% contribution (Punte, Gota & Peng 2013).

The classic focus on city planning does not include goods (Sjöstedt 2007) and most of the facilities or infrastructure in cities are developed, considering the requirements of the passenger transportation system. There is a strong need for coordination between passenger and freight transportation planning (Anderson, Allen & Browne 2005; Lindholm 2012; Taniguchi & Thompson 2018). Increasing attention towards UFT in the past few decades is because of growing awareness and concern about the environmental impact of transport and the implications for the economic vitality of cities caused by congestion problems (Lindholm & Behrends 2012). In order to improve the logistics performance and reduce the negative environmental or socio-economic impacts, various urban freight transport measures have been introduced in different cities across the globe, such as off-peak hour deliveries, urban consolidation centres, dedicated transhipment areas and loading or unloading zones, low emission zones, deliveries by electric vehicles and many more (De Oliveira et al. 2021; Karakikes & Nathanail 2020; Transmodal 2012; Taniguchi & Thompson 2018). The results, however, are disappointing, showing unwanted side effects or dependency on government subsidies, which is mainly because the measures were planned without taking into account the unique characterisation of the city and traffic, as well as the interests of all concerned stakeholders (Amaya, Arellana & Delgado-Lindeman 2020; Macharis & Melo 2011; Transmodal 2012). Therefore, a prerequisite for the successful implementation of urban logistics measures requires a good knowledge about the framework conditions of the particular context. A systematic approach for understanding the present city logistics scenario and then searching for the scope of improvement will help achieve the goal of sustainable urban freight transportation (SUFT).

Transferability to facilitate the application of sustainability measures has limitations because of a lack of data and literature. Therefore, there is a need for empirical research of specific megacities (Kin, Verlinde & Macharis 2017). Successful sustainable freight measure requires a regular and detailed study of traffic, including heavy commercial vehicles and business establishments (Oskarbski & Kaszubowski 2018). This study tries to understand the urban freight traffic characteristics in two cities and analyses their similarities and differences. Two cities, Gothenburg (Sweden) and Delhi (India), having significant socio-economic differences were selected to answer the following research questions:

This study focuses on characteristics of UFT (sector-specific contribution, trip characteristics, vehicular characteristics, freight delivery patterns, etc.) and also quantifies the negative impacts in terms of external cost (congestion, emissions, accidents, noise, climate change and infrastructure damage.), that is, cost borne by the society and not by the transport user (Sen, Tiwari & Upadhyay 2010). The study also tries to explain the relative contribution and performance of different sectors in urban freight traffic. The freight trip attraction (FTA) models are also developed, which can help in the estimation of total freight trips. Furthermore, it provides an overview of urban freight traffic characteristics of two different cities. The study findings will also be helpful for practitioners and policymakers to understand urban freight characteristics and associated problems, which will assist them in selecting appropriate sustainable urban freight measures for their city.

This study is organised as follows: Section ‘Case description’ provides the description of the cases; Section ‘Research approach’ discusses the research approach of data collection, data processing and data analysis; Section ‘Findings of the study’ deliberates on findings of the study focusing on the characteristics of freight traffic in the study zones; Section ‘Freight delivery patterns (based on receivers’ data)’ discusses freight delivery patterns and FTA models; Section ‘Externalities due to urban freight traffic in study zones’ deals with externality because of freight traffic in study zones, and Section ‘Conclusion’ presents conclusions.

Two commercial activity zones located in Gothenburg (Sweden) and Delhi (India) were selected for this study. Gothenburg is the second-largest city in Sweden, with around 0.58 million (2020) and spread over 450 km² area (Statistics Sweden 2020). Most of the commercial activities in Gothenburg are concentrated in the city centre area. In comparison, Delhi is India’s capital city, spread over 1500 km², with an estimated urban population of 19.3 million (Census 2011 2020). Delhi is a central distribution hub, where more than half of the freight imported is again re-exported. The central part of the city, Old Delhi, is very congested, which is a hub for wholesale and retail businesses. Commercial areas in Delhi are not concentrated, and they are spread all over the city. The major shopping districts in Delhi include Connaught Place, Chandni Chowk, Nehru Place, Sarojini Nagar and many other shopping malls in the city.

The first study zone (service area: 90 000 m²) is in Inom Vallgraven, inside the city centre of Gothenburg, Sweden. The study zone has a high number of retail stores, restaurants, hairdressers, supermarkets and offices, where business functions can occupy up to 95% of the area (Kaczorowska 2014). The study area covers eight entrance and exit streets and four loading bays. Site map of the study zone at the city centre of Gothenburg is shown in Figure 1a.

The area selected for the case study in Delhi is Sarojini Nagar Market (service area: 60 000 m²). It is a famous shopping attraction centre in Delhi, which witnesses thousands of daily footfalls, especially for clothes shopping. The market area consists of approximately 500 stores of clothes, footwear, electronics, furniture, restaurants, food outlets and groceries. Narrow streets and heavy pedestrian movement are some of the unique features of this market area. Figure 1b shows the site map of the study zone in Delhi.

Urban freight transportation is an emerging field of study where the ‘case research’ approach will be helpful to understand the characteristics and effects (Meredith 1998; Voss, Tsikriktsis & Frohlich 2012). The case study uses a data triangulation approach (use of multiple sources and methods), which improves the validity and reliability of case study findings. McCutcheon et al. (1993) have also mentioned that data collection may be in the form of personal observations, structured and unstructured interviews, review of previously collected data, status reports, etc. This study uses cordon counts (traffic data), retailers’ interviews (receivers’ data) and freight vehicle drivers’ interviews.

For traffic data, a secondary data source compiled by DB Schenker under the Send Smart Project during 18 and 19 December 2013 from 6:30 to 18:00 was used for the study zone in Gothenburg. However, traffic data for Sarojini Nagar market, Delhi, were collected on 21 August 2014 from 6:30 to 16:00 and 21:00 to 22:30 (vehicle entry restriction from 16:00 to 21:00). The vehicle registration plate number, time and place of entry or exit, company name and type of vehicle were recorded on each entry or exit point. The data were further processed using the available public databases, wherein vehicle-related attributes, such as vehicle ownership, gross vehicle weight, environmental class and fuel type, and the associated sector (public database for associated sectors is only available for Gothenburg), were identified (Delhi Vahan Sewa 2015; Transport Styrelesn 2014; UC All Companies Ltd. 2014). The traffic data were converted into trips using the following assumptions: (1) considering the multiple serving locations in the study or /nearby area, multiple entry or exit by the vehicle having a licence plate within the duration of 30 min was counted as a trip; (2) vehicles with a travel time less than 2 min were considered as through traffic and are not counted as destined trips; and (3) in the case of non-motorised vehicles, which do not have a licence plate (challenging to differentiate vehicles), single entry-exit of a vehicle is counted as a trip. Correction factors of 0.78 and 0.66 were applied for Gothenburg and Delhi, respectively, for unidentified trips, because of errors in data recording.

Furthermore, to understand the freight delivery pattern and to develop FTA models, personal interviews were conducted for 50 stores in Gothenburg and 102 stores in Delhi using a customised survey format (Holguin-Veras et al. 2012). During the survey, information regarding the type of activity, freight delivery characteristics such as type of vehicle, frequency of delivery and cargo weight, were also recorded. In the case of Delhi, retailers’ data were also supplemented by 68 drivers’ interviews, having the focus on trip origin destination, frequency of visit to the study area, type of visitor: regular or occasional, load factor (carried weight divided by permissible weight), number of stores served in a trip, nearby places served in a trip and the like.

This section discusses freight traffic’s general characteristics, such as freight trips and their distribution over time, share of different sectors in freight traffic, types of vehicles used and associated environmental emissions.

Based on random selection, data for freight deliveries were collected in the study zone at Gothenburg and Delhi. From the analysis, it was observed that only 27% of stores were part of any chain or franchise in Delhi; however, this share was high up to 46% for Gothenburg. It was found that the frequency of freight delivery trips is almost similar for both the study zones. More than half of the delivery trips are made on a weekly basis. The share of such trips in Gothenburg and Delhi was 56.6% and 58.3%, respectively. About 23.7% of delivery trips in Gothenburg and 27.2% of delivery trips in Delhi were made on a daily basis. The share of bi-monthly and monthly delivery trips ranges between 7% and 12%.

If we focus on the sector-specific contribution in FTA, it is observed that stores involved in the business related to HoReCa are attracting more daily delivery trips in both the study zones. Restaurants and Cafes in Gothenburg attract average 1.2 trips per day; however, the number of such trips in case of Sarojini Nagar Market is 1.7. It was also observed that food outlets in Delhi, having a relatively less floor area when compared with restaurants and cafes, are attracting more average delivery trips (2.3 trips/day). The number of service-related trips was high for Gothenburg, almost every store gets laundry, windows and carpet cleaning services on a bi-monthly basis, but a very few stores get laundry and other regular services in Sarojini Nagar Market. Around half of the deliveries in both zones were made by LCV/Mini-LCV. Low-carrying capacity vehicles like cycle and two-wheeler, etc., have a higher share of delivery trips to restaurants and food outlets than other stores.

For freight trip generation studies, linear regression models outperform constant trip rate models (Holguín-Veras & Jaller 2014). The efficiency of a model depends on the type of classification of sectors: land use-based, activity-based, industrial classification, etc. The classification should be such that it leads to homogenous classes regarding the determinants and pattern of FTG activity (Lawson et al. 2012). The activity-based classification was used to categorise different stores, such as retail: clothing or footwear, hotel and restaurants, and food outlets, etc. In order to predict the number of daily trips for different stores, FTA models based on employment size (no. of employees working in the shop), gross floor area, type of store, that is, part of a chain or not, amount of cargo (kg/week) were explored. It was observed that for FTA models, cargo weight was not a significant variable in the case of Sarojini Nagar Market. However, the floor area was found significant for food outlets, footwear and electronics stores. In the case of Gothenburg, a good linear correlation was observed between attracted freight trips and floor area, cargo weight and employment size, ranging from 0.45 to 0.80. For both the study zones, it is observed that the number of delivery trips does not depend on whether the commercial establishment is part of a chain or franchise.

Single-variable-based models gave higher precision in FTA predictions. The precision of area and employment-size-based FTA models for Sarojini Nagar market was found to be low. It was clear from the scatter plot of the data that there is not much variation in delivery trips because of change in floor area or employment size for such stores. However, it was observed that with an increase in floor area, the number of delivery trips increases up to an extent. However, relatively bigger size stores attract a lesser number of delivery trips than the small stores. It may be because of the availability of more storage area and the preference for consolidated deliveries. FTA models for different sectors are summarised in Table 2. The explanatory power of the model (R²) and t-statistics (two tail) with respect to the confidence interval of 95% are also provided.

From the FTA model summary in Table 2, it can be noted that with the exception of food outlets, all the coefficients for the area, employment size and delivery weight of cargo are positive for all the sectors. Hence, it can be concluded that with an increase in the value of these variables, the number of daily delivery trips would increase, whereas with the increase in floor area for food outlets in Delhi, the number of daily trips decreases. The number of delivery trips increases with the increase in employment per floor area for the electronics sector. This is also a proxy indicator of footfall and sales. Higher employment or floor area denotes that more employees are required per customer or sale.

In order to quantify the negative impacts of urban freight traffic in study zones at Gothenburg and Delhi, five types of externalities, that is, (1) congestion, (2) climate change, (3) accidents, (4) noise and (5) environmental emissions, were considered in our analysis.

In this study, the environmental externality covers the cost to the society because of freight vehicle emissions of four major pollutants, that is, hydrogen compounds (HCs), carbon monoxide (CO), nitrogen oxides (NO_x) and particulate matter-2.5 microns (PM-2.5). The vehicular emission (gm/km) is the function of vehicle type, environmental class, fuel type and traffic condition (ARAI 2008; Hausberger 2014). Total vehicular emission depends on trip length, for which data were collected through drivers’ interviews. The estimated amount of pollutants generated because of freight trips in the study zones is summarised in Table 3.

It is somewhat difficult to compare the negative impacts of freight traffic in both zones because of marginal external cost (MEC) functions for various externalities depending on the country’s economics. Because of the difference in traffic characteristics of both the cities and the unavailability of data, it was desirable to use the same marginal cost functions. Hence, MEC values, recommended by European Union (MOVE 2014) and updated to 2020 prices (Triami Media BV 2020), were used for Gothenburg, whereas MEC values were derived for freight vehicles in Delhi (Maibach et al. 2008; Sen et al. 2010). The approaches used for the derivation of MEC functions for Delhi were almost similar to the techniques used by MOVE (2014). However, it is to be noted that MEC functions for Delhi, like MEC_accidents, does not include risk elasticity and MEC_noise does not focus on health impacts. MEC_congestion for Gothenburg is based on the aggregated approach of FORGE model, whereas it is based on the value of time approach for Delhi (Sen et al. 2010; MOVE 2014). The MEC values for Delhi (2020 prices) are presented in Table 1. Climate change externality in this study only accounts for CO₂ emissions. The cost of CO₂ as EUR 110/ton for Sweden and EUR 70/ton for India was considered for calculations (Ministry of Finance 2020; Ricke et al. 2018).

Generally, CNG vehicles generate higher emissions of hydrocarbons and carbon monoxide when compared with diesel or petrol vehicles; however, emissions of particulate matter, nitrogen oxide and carbon dioxide are lower from CNG vehicles (Blanchard, Teil & Chevreuil 2006; Lowell 2013). The results reveal that CO₂ emissions generated by freight vehicles in Delhi were almost half of Gothenburg. However, emissions of hydrocarbons are much higher, which is more damaging to the climate than CO₂ (EIA 2014). There is a difference in the characteristics of vehicles and fuel type used for both the cities. Although CNG is the fuel for most freight vehicular trips, relatively outdated technology (BS-I to IV) is in place for Indian vehicles. This is responsible for higher PM emissions. The concentration of PM_2.5 in the air has reached an alarming level for Delhi, and the results of this study also show that emissions of PM_2.5 from freight activities in Delhi are almost seven times when compared with Gothenburg, from the almost similar amount of freight activities (Beig et al. 2020).

A comparison of external cost for both the study zones is shown in Table 3. For conversion of Rupee into Euro, a conversion factor of 1 Euro = 85 Rupees (2020 prices) is used. Social cost because of congestion, environmental emissions and noise are less in Delhi because the cost for health damage, time loss, injuries, fatalities, ecological damage, etc. is also less. Therefore, comparing the external cost values is not a realistic approach for the measurement of negative impacts because of the unequal scale of conversion of externality to monetary value. However, it is interesting to note that India as a low-income country, where the value of statistical life and damage cost of accidents is much lower when compared with high-income groups, external cost because of accidents is much higher than developed and high-income countries like Sweden (Miller 2000). This indicates that fatal crash rates of freight vehicles in urban areas are much higher in Delhi (Mohan et al. 2009) when compared with Gothenburg, and that is why MEC of accidents for freight vehicles is also high.

Results of the sectoral analysis for Gothenburg reveal that 3PLP is the dominant sector and contributes one-third to total external cost. HoReCa is the second-highest contributor. The share of express courier and post sector is lowest. At the same time, other sectors have almost equal impacts. The share of congestion externality is far higher than other externalities. The share of accident externality is lowest in total external cost, and 3PLP is the dominating sector under this category too. The 3PLP and the construction and maintenance sectors are more responsible than any other sector for noise-related external costs because of the use of big-size trucks and most of the fleet operations during peak hours.

With the case study of Gothenburg and Delhi, this article presents the urban freight scenario in commercial land use. The findings of the study will be helpful for further research studies on sustainable freight movement strategies. It is observed that most of the freight movement in both the study zones are with small-sized vehicles, and their share ranges between 75% and 92%. The total externality due to freight traffic can be reduced with an improved load factor. A variety of vehicles, even passenger vehicles like autorickshaws, cycles and two-wheelers with a substantial share of 28%, are used for freight movement in Delhi. In both the study zones, more than 50% of the deliveries are made weekly, and more than 90% of the delivery trips are running on conventional fuel. Almost 74% of freight vehicle drivers were regular visitors to Sarojini Nagar market, which implies the presence of an informally organised sector and the scope of better management by policy-based measures for sustainable urban freight.

Although 76% of freight trips in the study zone at Delhi were made by CNG vehicles, total PM emissions were manifolds in comparison with Gothenburg. The vehicles meeting BS-IV/Euro-IV and below emission norms generate a significantly higher amount of PM emissions than those meeting Euro-V and above emission norms (Crippa et al. 2016). Accident rates for freight vehicles in urban areas are much higher in Delhi when compared with Gothenburg.

The findings of this study indicate substantial dissimilarities in freight traffic characteristics and operations in Gothenburg and Delhi. City-specific research is needed before adopting sustainable urban freight measures to improve the transportation system’s performance and reduce negative impacts. Furthermore, most of the variables used in FTA modelling in this study have shown relatively low predictability, and there is a scope for future research to identify more significant variables for higher accuracy FTA models.