Reducing fuel consumption and carbon emissions through eco-drive training

Transportation Research Part F 46 (2017) 96–110

Contents lists available at ScienceDirect

Transportation Research Part F journal homepage: www.elsevier.com/locate/trf

Reducing fuel consumption and carbon emissions through eco-drive training Kemal Ayyildiz a, Federico Cavallaro b, Silvio Nocera b,⇑, Ralf Willenbrock a a b

Deutsche Telekom RO China, Beijing Lufthansa Center, C213, CN-100125 Beijing, China IUAV University of Venice, Santa Croce 191, I-30135 Venice, Italy

a r t i c l e

i n f o

Article history: Received 28 July 2016 Received in revised form 17 December 2016 Accepted 20 January 2017

Keywords: Eco-driving CO2 emissions Freight transport Field trial China

a b s t r a c t Freight transport is responsible for about 45% of total CO2 emissions caused by mobility. To reduce its contribution without limiting the quantity of goods distributed, the technological improvement of vehicles and alternative fuels is only a partial solution. Indeed, it should be flanked by other integrative push- and pull-measures implemented at the policy level. Among them, eco-driving represents an option that covers strategic, tactical and operational decisions and provides suggestions to drivers, as well as real-time monitoring of their performance. This paper brings some evidence to such assertion, first providing a summary of the literature assessing the impacts of eco-driving, and then highlighting the limited amount of studies regarding freight transport. The results of an eco-driving field trial conducted in the Chinese province of Jiangsu, whose data were gathered using a method based on real-time normalized indicators specifically developed by Deutsche Telekom AG, are then presented. 15 heavy-duty and 10 light commercial vehicles were monitored over a period of four months, and information was collected on more than 5200 trips, covering a total distance of 439,000 km. A comparison between the driving styles before and after the training revealed a reduction of unitary fuel consumption for heavy-duty vehicles (
5.5%), while no significant variations were visible for light commercial vehicles. The application of this research method also yielded useful information about braking, acceleration and standstill, which are normally not considered in these types of evaluations, but can be highly valuable to drivers who wish to modify their behaviour towards a more efficient style of driving. Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction Greenhouse gases (GHGs) are a relevant indicator to evaluate transport sustainability (Black, 2010). In relative terms, transport counts for about 30% of total anthropogenic GHG emissions and its impact has increased by about 22% in comparison to 1990 levels, thus revealing a critical environmental issue that needs to be addressed urgently (EU, 2014). As it accounts for almost 45% of total energy that is consumed by transport (Sims et al., 2014), freight transport is significantly co-responsible for this rise. Even if the adoption of alternative fuels can grant substantial benefits in terms of Tank-To-Wheel emissions (Nocera & Cavallaro, 2016a), the technological development of vehicles and alternative fuels is not enough to curb transport GHGs ⇑ Corresponding author at: Department of Architecture and Arts, IUAV University of Venice, Dorsoduro 2206, I-30123 Venice, Italy. E-mail address: [email protected] (S. Nocera). http://dx.doi.org/10.1016/j.trf.2017.01.006 1369-8478/Ó 2017 Elsevier Ltd. All rights reserved.

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

97

on its own. Specific policies to encourage a modal shift towards less polluting systems are necessary (Dray, Schäfer, & BenAkiva, 2012), as well as the definition of the correct role assigned to GHGs within urban mobility plans (Nocera & Cavallaro, 2014). Policy-makers are aware of this condition and, since the early 1990s, the EU has constantly increased its efforts to create a framework that also includes this aspect (DGET & Transport, 2006). The mid- and long-term continental goals are coherent with this approach, aiming at reducing GHG emissions by 40% before 2030 (EC, 2015) and by around 60% before 2050 (EC, 2011). At a global level, the agreement achieved at the Paris climate conference (UN, 2015) is also coherent with this necessity. Stimulated by these international climate conventions, even private stakeholders (such as large logistics enterprises) have announced voluntary GHG emission reduction targets, which can be obtained through the adoption of specific transport measures. They are known as ‘‘push-” and ‘‘pull-” measures: the former are imposed on freight operators to obtain more equitable transport pricing, seeking to require transport users to bear a greater proportion of the real travel costs. They include financial instruments (e.g., taxes, charges and tolls) and technical and regulatory constraints (e.g., orders and bans). Pull-measures are implemented in order to minimize the impacts of private trucks and commercial vehicles by improving the attractiveness of existing less polluting alternatives. The results deriving from the adoption of these measures are monitored through annual reports, which highlight the fuel costs registered for each vehicle. Indirectly, they can provide the approximate amount of GHG emitted and the negative impacts upon the global climate balance caused by logistics operations. This approach is part of the so-called macromodels, as it tries to explain how changes at a large temporal and spatial scale influence carbon production. However, these reports do not give evidence of any positive investment in low emission vehicles and of the positive impact of specific measures undertaken at the unitary level, which contribute to reduce the relative figures of GHG emitted per ton of load and mileage. The microeconomic evidence suggests the importance of considering the individual heterogeneity and decisions taken at the individual level for traffic and modal split outcomes (so-called ‘‘individual behaviour”). To this aim, highly accurate methodologies for detecting fuel available are still present, such as CAN Bus fuel detection technology, injection pipe sensors and sensors inside the tank measuring the filling level. Accurate telematics devices are able to combine these fuel measurements with the position detected from satellite receivers sending acquired telecommunication technology to an external server to enable on trip and post-trip fuel consumption monitoring. However, the overall operational costs of such systems are still too high and cheaper solutions have to be implemented to grant diffusion on a broader scale. This paper assesses the potentialities of a specific transport measure (eco-driving) and its implications towards the reduction of fuel consumption and carbon dioxide (CO2) emissions from freight transport. CO2 is a valid indicator to assess global warming caused by freight (McKinnon & Piecyk, 2009), since it counts for 93–95% of total GHG emissions from freight transport. Results are provided by integrating a micro-modelling approach to the annual reports previously recalled. Our intention is to assess the reliability of this model in providing real-time indications about driving styles and consequences on fuel consumption. At the same time, the efficiency of eco-driving as a measure to reduce CO2 emissions caused by freight transport is analysed. The paper is structured as follows: Section 2 explains the nature of eco-driving, including a literary review of its effects, the phase of implementation and the expected fuel and carbon potentiality. Section 3 describes the methodology that we adopt to measure the expected reduction of fuel consumption and to help drivers know how to improve their driving efficiency. This methodology is tested on a case study in China and discussed in comparison with results found in the literature (Section 4). Some final remarks and policy implications (Section 5) conclude this contribution. 2. Eco-driving, fuel consumption and CO2 emissions Eco-driving is a generic term used to describe an energy-efficient use of vehicles that is based on the decisions and behaviours adopted by drivers. Behind this concept, it is assumed that while the performance of cars has improved rapidly due to technological developments and the introduction of alternative fuels, drivers have neither modified nor adapted their driving behaviours. If an adequate education about strategic, tactical and operational decisions is provided to drivers, eco-driving can contribute to limit overall fuel consumption and CO2 emissions (Sivak & Schoettle, 2012). More in detail, specific measures can be adopted in the pre-trip, on-trip and post-trip phases (Wengraf, 2012). Regarding pre-trip activities, eco-drive manuals suggest performing maintenance checks of the vehicle, gauging the tyre pressure and assessing if there is any unnecessary vehicle weight, and making a detailed plan of the trip that seeks to find short-distance alternatives. During the trip, some practical recommendations may be useful, including a limited use of air conditioning, the avoidance of unnecessary engine idling, sharp acceleration and heavy braking, changing up gear as soon as possible and anticipating traffic behaviour. Finally, post-trip activities include reviewing data about the trip, which can be made by employing specific apps or satnavs. Eco-driving does not imply the use of all these recommendations. However, some of them can be used in a combined form to obtain multiple benefits, including private financial aspects (the saving of fuel and money), safety (less accidents), environment (reduction of global emissions, local air pollutants and noise), and society (more responsible driving stile and reduction of traffic congestion). In the most commonly used form of eco-driving measures, drivers are given advice in classes or training sessions. The organizers measure differences in fuel consumption and CO2 emissions before and after the training, thus determining the efficiency of the lessons. The results are then communicated to drivers using two main forms of feedback (Fig. 1). The real-time indicator relates the driving style of the drivers to other external conditions, such as type of road or congestion

98

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

Tool

Feedback Real-Time Indicator Instantaneous fuel economy or engine power

Eco-driving Trip End Summary - Total fuel - CO2 emissions - Costs of driving

Behavioural changes

Results

Outcomes

Driving behaviours - Shift gear sooner - Maintain steady speed - Accelerate softly - Decelerate smoothly - Turn off engine

Smoother drive; less unnecessary idling

Travel behaviours - Trip generation - Trip chaining - Alternative destinations - Mode shift - Alternative routes

Reduced fuel usage Reduced GHG emissions

Reduced number of trips Reduced VMT

Fig. 1. Communication of eco-driving results, conceptual framework. Source: Boriboonsomsin et al. (2010)

in order to determine the fuel economy of the vehicles. The instantaneous feedback received by the drivers allows them to adjust their driving behaviours during the trip. Conversely, the trip end summary illustrates to drivers their actual costs incurred by driving on a trip-by-trip basis, thus suggesting changes to their driving style, but only once that the trip is completed. Antonissen et al. (2013) state that both of these forms of eco-driving measures could contribute to a reduction of consumption, but with distinctions. On-trip eco-drive support is more efficient (reduction higher than 10%), while post-trip feedback is expected to generate reduction lower than 5%. However, the former is more expensive and needs a specific real-time technology, which in most cases requires customization in order to achieve the desired results, while the latter can be provided also using simple methods, such as analysing the fuel cost data. According to this method, sustainability reports are generated from financial booking systems taking the overall fuel costs and the average load to sum up l/(t ⁄ km). These figures do not allow any fuel reduction measures on-trip or short-notice post-trip, as all trip details get lost in between the two moments where fuel is pumped into the tank. Accordingly, the effects of eco-drive training cannot be properly assessed. Results in terms of fuel savings and reduction of CO2 emissions (normally expressed as a variation in comparison to a baseline scenario) are in most cases difficult to compare. Indeed, they are based on different methodologies, and consider a huge variety of vehicles and participants, which circulate along road types with different topography and congestion patterns. Alam and McNabola (2014) provided an overview of eco-drive measures assessed in scientific articles. The range of fuel and CO2 reduction derived from their observations spans from 1% to 34%. Particularly, results seem largely affected by the nature of the study (e.g., field trial or simulation), the type of traffic considered and the specific eco-drive measures adopted. Rather than a comprehensive evaluation of the effectiveness of eco-driving, it seems more reliable to assess the single studies, according to the transport mode considered. As far as car drivers are concerned, Boriboonsomsin, Vu, and Barth (2010) analysed the behavioural changes of 20 self-selected samples of drivers in Southern California, showing contradictory results. The fuel economy on city streets ranges from
5% to +24% (with an average improvement of 6%), while the fuel economy on highways varies from
12% to +13%, with an average improvement of 1%. This would mean that in some cases ecodriving had even negative effects in terms of fuel savings. These results are partially consistent with the work of Beusen et al. (2009): their analysis, based on 10 drivers, revealed a generalized reduction of fuel consumption by about 6% (but in some cases, no saving of fuel was registered). Qian, Chung, and Horiguchi (2013) limited their analysis to an urban road (1 km) with three intersections, by evaluating the consequences on the fuel consumption of 15 drivers. A generalized reduction was registered, varying from 3% to 19%. Jeffreys, Graves, and Roth (in press) assessed the driving behaviour, of 1056 private drivers in Australia. The driver education led to a statistically significant reduction in fuel (0.51 l/100 km,
4.6%), with an annual reduction in CO2 emissions of 169 kg per vehicle. Baric´, Zovak, and Periša (2013) tested the effects of an ecodriving campaign in Croatia, by measuring the fuel consumption of a commercial vehicle prior to education, immediately following and three months after the eco-training. Results deriving from the single vehicle are extended to a fleet of 1500 vehicles through a simulation, causing an expected reduction of CO2 emissions by 32%. Finally, Wu, Zhao, and Rong (2015) assessed the effectiveness of eco-driving trials on taxi drivers in Beijing. A reduction of fuel consumption equal to 19% was registered in the first month, reduced to 14.5% after 4 months. Eco-driving efficiency was also tested for bus drivers. Zarkadoula, Zoidis, and Tritopoulou (2007) measured an average saving of 4.35% in fuel consumption, even if the number of participants in their field trial was very limited (3 bus drivers) and only a 15 km route was considered. These results are in line with those provided by Rolim, Baptista, Duarte, and Farias (2014), who tested the effects of an eco-drive education session on bus drivers in Portugal. Measurements indicate an average 4.8% fuel consumption decrease, corresponding to CO2 savings of 6.56 g/km. Strömberg and Karlsson (2013)

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

99

assessed the impacts of two different eco-driving programmes, with no significant differences in absolute terms: in both cases, a 6.8% saving of fuel and large decreases in instances of harsh deceleration and speeding were found. Sullman, Dorn, and Niemi (2015) tested a measure of eco-driving on bus drivers adopting a simulation approach. Fuel economy for the treatment group improved significantly immediately after the eco-drive training (11.6%) and this improvement was even larger six months after the training (16.9%). However, results of the simulation reveal only minor differences in terms of CO2 emissions, which are very difficult to explain and raise doubts about the reliability of the method adopted. Finally, regarding Heavy-Duty Vehicles (HDVs), the scientific literature is more limited. Symmons and Rose (2009) led a test in Australia in which 15 HDV drivers underwent an eco-driving course. Their driving was assessed for various eco-drive variables as they completed an 18-mile circuit in normal traffic immediately after the course and again 6 and 12 weeks later. Compared to pre-course measures, these drivers reduced their fuel consumption by an average of 27%, which is a value very high in comparison to the other studies previously recalled. Other studies about the effectiveness of eco-drive training on HDV drivers are not provided, thus making it difficult to evaluate its effectiveness for freight transportation. To address this lack, in the next sections we present the results of an assessment by Deutsche Telekom AG about an eco-drive training for HDV drivers made in the Chinese province of Jiangsu. 3. Methodology 3.1. Theoretical background The method presented in this section to evaluate the effectiveness of eco-driving on freight transport adopts a micro-scale approach. This means that the real-world driving patterns of the single vehicles and the drivers’ individual behaviours are considered, which affect the transport demand, the modal split, the route choice and the trip timing (Samaras et al., 2012). This approach implies an analysis of the individual vehicle fuel consumption, stops, speed, acceleration, deceleration and engine power to model instantaneous consumption of fuel (Demir, Bektasß, & Lavorate, 2011). Operatively, the physics of driving (Robert Bosch GmbH, 2014) include five contributors to calculate the total energy needed to let a vehicle circulate: energy needed to overcome aerodynamic resistance (Eair), rolling friction resistance (Eroll), inertia by acceleration (Eacc), slope resistance driving uphill (Eg), energy needed to keep the engine running in idle mode (Eid). While acceleration, rolling friction and gradient are mass dependent, aerodynamic and standstill are mass independent. The five components can be used to model the absolute and relative fuel consumption of a vehicle under specific driving conditions, as in formula (1).

Etot ¼ Eair þ Eroll þ Eacc þ Eg þ Eid

ð1Þ

The aggregate fuel consumption measured before and after a specific timeframe, deriving from fuel sold, is broken up to these five components, which are measured independently. This makes the calculation more accurate, allowing the real comprehension of the components that mostly affect the fuel consumption, which is not possible by adopting the aggregate methodology. All these energy contributors are related to the detection of speed profiles per second and compared to standard driving cycles (e.g. UNECE R101; UN, 2012). By calculating the % deviation per physical energy, each trip of a truck can be classified according to the instantaneous ± value relative to the driving cycle. To this aim, the following steps are required:  Identification of the type of trip (urban, extra urban or highway);  Calculation of the normalized fuel consumption in a defined time window for the cycle selected, according to the constant parameters;  Calculation of the real speed profile;  Comparison between normalized driving cycle and real speed profile in the same timeframe for all the single energy elements (acceleration and deceleration, rolling and aerodynamic resistance, standstill and energy value due to driving uphill and downhill);  Classification of a trip and highlights of particular events, distinguishing them according to external road conditions and individual driving behaviour. Once the single components are defined, data has to be communicated to the drivers in a way that allows them to understand the reasons for high or low fuel consumption and adapting their individual driving behaviour according to external road conditions or purely psychological reactions. These values have to be expressed both in absolute and relative terms. Among the former, information about costs, fuel used, CO2 emissions, room for improvement and other practical information should be provided. Referring to the relative values, Key Performance Indicators (KPIs) normalize the values according to the single ton transported and the unitary distance. KPIs can be expressed in several ways; when the efficiency of eco-drive training is to be assessed, Energy Performance Indicator (EPI; see formula (2)), and Acceleration Performance Indicator (API, see formula (3)) can be adopted. Both the indicators consider the unitary stretch of 1 km. API refers mostly to the Eacc component of formula (1), while EPI is more comprehensive and includes other components, such as the stand still and the braking indexes. More in detail, EPI, expressed in cl/(t ⁄ km), indicates the average quantity of fuel required to transport 1 ton of freight. API designates the energy consumption necessary to get a vehicle up to speed or to increase from a constant to a

100

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

higher speed. This indicator can be expressed either in additional litres (or their submultiple) of fuel or in additional energy required (kW h). In the first case and referring to unitary values, acceleration might provide small integer numbers. Therefore, we decide to express API in kW h/(t ⁄ km), which allows us to get integer ranges of similar number scale as for EPI.

EPI ¼

  Av g Fuel centiliter in Mass  1000 t  km

ð2Þ

API ¼

  Acceleration kW h in Mass  Distance t  km

ð3Þ

With such information, drivers can be informed about instantaneous fuel consumption and driving style, adapting their driving behaviour accordingly. On parallel, more strategic information about the managerial aspects of the vehicular fleet has to be provided to the dispatchers, including the aggregate values about trip duration, distance covered, average speed, maps with the details of the trips and the road segments where fuel consumptions are abnormal or unexpected. This data is useful for the ex-post evaluations, as well as for the activities related to the route planning and the rescheduling of the deliveries: recalling Fig. 1), these aspects belong to the trip end summary. 3.2. Low carbon mobility management cloud platform architecture To obtain information about the KPIs, Telekom AG has developed a low-cost solution called Low Carbon Mobility Management Cloud Platform Architecture, which considers the economic constraints felt by most logistics companies and their difficulties in investing in the introduction of new technologies. A Plug-&-play solution is adopted, where GPS data provided by the HDVs is sent to a database, elaborated, analysed, mapped and then given back in a simplified version to drivers and dispatchers. A schematic architectural representation of the platform is presented in Fig. 2. The process goes as follows: primary GPS data about vehicles derives from a mobile phone installed in the truck. Such data, which includes time, latitude, longitude, speed and altitude, are transmitted every fifteen seconds to the server platform. Since the average duration of a mobile phone with GPS turned on is four hours, which is a short period in comparison to the average daily tour of a truck, the mobile phone is equipped with an external battery that extends the duration of the battery to the duration of the entire trip. With this system, the driver has to turn on or off the device only once a day and can use the smart phone as telephone or mobile Internet device during the day without losing any information. Before the beginning of each trip, all users must register their vehicle configuration according to a predefined database (more than 8000 types are available) and user information and insert the payload of the specific trip. Data is transmitted to the Low Carbon Mobility Management (LCMM) platform, which includes three main components: the MySQL database, the PHP code server side and the android application. The collected GPS data is analysed and evaluated within the data base server using PHP programing language. Inside the server, the percentage deviation relative to the urban and the extra urban reference cycle is calculated and used as a basis for determining fuel consumption and CO2 emissions.

Fig. 2. Low carbon mobility management cloud platform architecture.

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

101

The feedback on driving performances is provided back to freight transport operators in two different ways: either through the website platform (for post-trip evaluations and tour planning of the fleet manager), or through the androidapp (for real-time evaluations). The website platform provides detailed information about the time of the trip listed as well as duration of trip, distance travelled, average speed, EPI and API (Fig. 3). Based on the consumption registered by the mobile phone, the fuel saved by the eco-drive behaviour and the performance of the driver can be determined, both in aggregate and disaggregate forms. They are visualized thanks to specific tools, such as summary sheets about driving performance of the single vehicles, their speed curves and an emission profile map. Compared to on-board unit platform solutions, many new features were introduced based on the physics of driving calculations, usually not available for freight transport operators, such as time lost, grade work, normalized braking and acceleration index. The android app includes different interfaces to monitor the real-time information about a trip. In Fig. 4, an example of three screens are presented. The speed panel pointer (Fig. 4, left) indicates the velocity variation status of the trip and the CO2 emission figure, which help the HDV driver to evaluate his eco-drive behaviour. Some important parameters (fuel saved, API, EPI, distance covered) are listed. More concisely, the trip cost interface (Fig. 4, middle) makes drivers aware of the trip cost estimate, CO2 emissions and room for improvement in terms of fuel consumption. Finally, the real fuel consumption screen (Fig. 4, right) monitors instantaneous driving performance. In the next section, the LCMM solution is tested to describe the effectiveness of eco-drive training in the Chinese province of Jiangsu. The aim is to understand whether the real-time component provides reliable information, which is an approach that can be considered consistent with the traditional macro-modelling approaches adopted to evaluate the CO2 efficiency of vehicular fleets. As previously mentioned, this would not only contribute to an ex-post evaluation of the performances, but could also grant a prompt modification of the driving style in order to reduce fuel consumptions and CO2 emissions. In this way, it is possible to understand the real potentialities granted by field trials, giving at the same time a practical solution to improve their effectiveness. 4. Eco–driving and freight CO2 emissions: evidence from a case study in China

4.1. Description of the Jiangsu area Jiangsu is a flat province located in the eastern part of the People’s Republic of China. Although the Province is rather small (it is the second smallest Province of the entire State), Jiangsu is the most densely populated and it has the second highest GDP of the whole nation after Guangdong. This derives from a dynamic economy, based on agriculture, light industries and trade. Coherently, several Export Processing Zones have been developed and freight transport constitutes one of the main pillars of the Provincial economy. Among the different freight operators in the territory, the Linsen Logistics group is a company with more than 1200 employees. Linsen decided to introduce private solutions to reduce their fuel consumption

Fig. 3. LCMM cloud platform architecture, example of a screen taken from the website platform.

102

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

Fig. 4. Low carbon mobility management cloud platform architecture, details of the app: speed panel, trip cost, real fuel.

and CO2 emissions from the operational phase. To this aim, Linsen has adopted a measure of eco-driving for their employees: if they save a part of the fuel (compared to that expected to be used for their deliveries), this part could be used as a bonus for their private use. A field trial for daily logistics operations was also provided. The real impact granted by this eco-drive measure is measured by the system described in Section 3. Preliminarily, a selected number of vehicles from their logistics fleet operator is equipped with GPS smart phone devices collecting trip data from the daily logistics operation of the company. 25 trucks belonging to three main categories have been selected in order to make the analysis as representative of the everyday realities of road transportation as possible. First, the HDV is a typical liquid tank container transporting dangerous liquid goods. The empty weight is 14,280 kg, maximum payload is 25,700 kg, and the weight of full load is 39,980 kg. The standard average fuel consumption is 35 l/100 km. The second group of trucks is constituted by Light Commercial Vehicles (LCVs), transporting chickenfeed for agricultural chicken farming. The empty weight of these trucks is 8545 kg, usual load weight is 7350 kg, and full load weight is 15,895 kg. The standard average fuel consumption is 22 l/100 km. The third category of vehicles is LCVs that transport live chickens. The empty weight of these vehicles is 6205 kg, the load is 9600 kg, and the full load weight is 15,805 kg. The standard average fuel consumption is 21 l/100 km. From the fleet operator and the vehicle model, the cross section of the truck size in length, width and height was known, thus yielding the aerodynamic resistance. The characteristics of the vehicles are summarized in Table 1. While the physical parameters are quite close for the different weight classes, a difference comes up with the full load per weight class. Here empty weights as well as half load and full load differ. The average load factor ranges from 50% to 80%, depending on the amount of empty and fully loaded trips that occurred between two fuel-filling events. 4.2. Results from the field trial All vehicles have been equipped with smart phones and external batteries. The average annual cost of the LCMM system installed on a single vehicle is estimated at 1200 yuan renminbi per year (¥/year, equal to €144/year1), which means an overall cost of ¥30,000/year (€3595/year). Drivers were trained in how to use the devices, without causing any disturbance and interference in their duties of delivering goods. An introduction of data privacy was given in writing, clearly specifying that when turning on the mobile phone, speed profile and vehicle position data were sent and made visible to an external

1

The EUR-¥ exchange rate at 31.12.2013 is fixed at 0.1198 (Exchangerates, 2013).

103

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110 Table 1 Characteristics of vehicles adopted in the eco-drive field trial. Truck type

Vehicles (n°)

Fuel

Empty weight (kg)

Payload (kg)

Total weight (kg)

Average fuel consumption (l/100 km)

Rolling friction coefficient (–)

Air drag coefficient (–)

Cross-section area (m2)

Engine efficiency (%)

Fuel value (kW h/l)

HDV LCV type 1 LCV type 2

1–10; 16–20 11–15 21–25

Diesel Diesel Diesel

14,280 8545 6205

25,700 7350 9600

39,980 15,895 15,805

35 22 21

0.008 0.008 0.008

0.8 0.7 0.7

9.25 8.80 8.80

0.65 0.48 0.48

9.72 9.72 9.72

Table 2 Reduction of fuel consumption before and after eco-drive training. Indicator

Unity of measure

LCVs

HDVs

Distance saved Monthly fuel saving Monthly fuel saving per vehicle Yearly fuel saving per vehicle Yearly fuel saving Fuel reduction Monthly economic saving Monthly economic saving per vehicle Yearly economic saving per vehicle Yearly economic saving

km l/month l/month l/year l/year % ¥/month ¥/month ¥/year ¥/year

57,330 126 13 151 1514 1.00% 971 97 1165 11,654

131,076 2724 182 2179 32,690 5.94% 20,976 1398 16,781 251,711

third-party server via Internet. Participation of all drivers was declared on a voluntary basis. It was also guaranteed that no names of drivers were directly mentioned, but only truck numbers from truck number 1 to truck number 25. For motivational purposes, ¥100 (€12) were given to all drivers joining the field trial and every two weeks another ¥200 (€24) more was granted for the top 3 drivers sending GPS trip data to the server. In the first phase (which lasted 2 months: October–November 2013), the performance of participating drivers was recorded, according to their initial driving style. Subsequently, in phase 2 (from December 2013 to January 2014) all participating drivers received professional training from an engineering company specialized in the field of eco-drive training and their driving performance was measured as well. The on-trip feedback was expected to ensure regular awareness of driving behaviour, allowing an adaptation of the driving style according to the main criticalities detected by the LCMM. A total number of 5500 trips (corresponding to about 475,000 km) have been recorded. Table 2 and Fig. 5 compare the average fuel consumption before and after the training, expressed in litre per 100 km. The values have been calculated as average values of all trips recorded during the time period before and after the training took place. Kilometres driven have been obtained from the odometers and then the average fuel consumed (expressed as litre per 100 km) has been derived. In referring to the reduction of fuel, positive values mean savings in fuel after the eco-drive training, while negative values indicate an increase in fuel consumption. 4 of the 15 HDVs (n° 01, 03, 04 and 17) did not show any improvements after the ecodrive training, whereas others had even double-digits% of savings, with a peak for HDVs n° 06, 07 and 19 (up to
32.68%). The reduction in fuel consumption seems to be less significant for the LCVs (n° 11–15 and 21–25), where the average value is lower than 1%, while for HDVs the fuel reduction is about 6%. In aggregate terms, a total 4% saving of fuel is registered. This type of evaluation includes neither payload, nor the length of the trip and the characteristics of the road network. Furthermore, the characteristics of the engine or the fluctuating aerodynamic and road surface conditions are not considered. All these oversimplifications, which are legitimate for a rough analysis, make the results not accurate enough for an ecodriving campaign. Indeed, the reduction of fuel consumption (as well as the efficiency of eco-driving) can be cancelled if the truck has higher payload after the eco-drive training. Similarly, the optimization of the load (which means reduction of empty trips) can lead to misleading results if not adequately taken into account. To overcome these criticalities, the LCMM system can provide useful integrative information to understand the real performance of vehicles, thanks to the KPI’s EPI and API (see Eqs. (2) and (3)). The interpretation of this data requires a caveat. Evaluation methodologies based on speed profile and real-time data collection can present two error sources. The former is related to the regularly changing payload influence, while the latter is linked to the engine efficiency parameter, which can change dramatically during the engine operation, especially during short trip delivery (the engine needs up to six times more fuel when it is cold). This can lead to a difference of up to 10% in comparison to data presented in Fig. 5, which makes it difficult to compare results. To reduce this difference, values derived from mileages shorter than 5 km, average speed values lower than 5 km/h and trips where the percentage of time at standstill is above 90% are not considered. Referring to the first source of error (changing payload influence), in Section 3.2 it was highlighted that before each trip drivers have to insert the characteristics of the vehicle and the payload. If they forget to insert this last value, the software assigns a payload equal to 0. This can obviously generate misleading values. These conditions occur rarely, constituting only less than 300 trips out of a

104

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

Fig. 5. Reduction of fuel consumption before and after eco-drive training.

total of 5500 travels monitored in the Jiangsu field trial. Expressed in km, 439,000 out of the 475,000 km driven (about 92%) have been correctly registered and constitute a reliable basis for the following analyses. To obtain normalized values, the LCMM uses the GPS input to measure mileage and speed, estimates an average payload and weight, fixed and stable physical conditions in aerodynamics and road surface and then sums up all this data. Referring to the influence of payload for the fuel and CO2 performance analysis of a given fleet of trucks, all recorded trips have been normalized by weight, estimating an average of 160% including 100% of empty vehicle weight plus 60% payload on top of it as average payload for all trips. This average value is derived from the freight weight as registered in specific booking systems of the transport company and it is coherent with other previous experiences regarding freight transport along transnational corridors (Cavallaro, Maino, & Morelli, 2013). Referring to HDVs (vehicles 1–10 and 16–20 of Fig. 5), Table 3 presents a summary of the main results deriving from the field trail in Jiangsu. A total number of 316,113 km has been validated: 196,656 km were registered before the eco-driving campaign and 119,457 km after the training. The absolute values of fuel consumption and CO2 emissions decrease respectively from 72,514 to 41,630 l and from 188,535 to 108,238 t. However, these figures are not significant per se, because they depend on the distance covered, which is not comparable. Indeed, average indicators provide more interesting results, revealing a generalized improvement in the driving behaviour. First, the unitary fuel consumption decreases from 0.369 to 0.348 l/km (
5.5%). By assuming a CO2 transformation coefficient of 2.6 kg CO2/l fuel, this determines a decrease also in the unitary CO2 emissions from 958.71 to 906.08 g/km. When analysing the KPIs, EPI shows a reduction by 8.8%, passing from 1.05 to 0.96 l/(100 km ⁄ t), whereas API decreases on average by 3.7%, from 2.06 to 1.98 kW h/(100 km ⁄ t). The braking index, which is the average of harsh braking deceleration, shows an improvement after the training, passing from
0.22 to
0.19 m/s2. Finally, the grade work energy (i.e. the extra energy required to drive uphill) and the standstill are calculated both as percentage of total energy consumption. As far as standstills are concerned, an increase of the values is visible, while grade work energy shows a slight decrease after the training. Referring to LCVs (vehicles 11–15 and 21–25 of Fig. 5), a more limited number of km has been registered: 122,917 km in total, of which 64,799 km have been recorded before and 58,118 after the eco-driving training (Table 4). Even in this case, absolute values regarding fuel consumption and CO2 emissions (which passes from 14,997 l and 38,993 kg to 13,345 l and 34,698 kg) are not significant. The focus should be limited to one on average and normalized values. Here, a reduced benefit deriving from the eco-driving training is visible: average fuel consumption is stable at 0.23 l/km; coherently the average CO2 emissions shows no significant variations, passing from about 602 to 597 g/km, with a modest reduction equal to 0.8%. The KPIs indicate positive results: API passes from 1.89 to 1.68 (
10.9%), EPI decreases from 1.37 to 1.30 (
4.8%), while braking index lowers from
0.27 to
0.20 m/s2. These results are then visualized through a radar chart that is based on a 1–5-progressive star range. Intervals are based on a statistical analysis of the quartile values derived from the data collection, calculated for each indicator separately. Table 5 indicates how the ranges are calculated. Fig. 6 represents the aggregate performances of HDVs (left) and LCVs (right), before and after the eco-drive training. Both HDV and LCV drivers (indicated with the violet and the green lines) show an improvement of the performances related to the normalized indicators after the eco-driving training. Particularly, a generalized increase of the score regarding fuel consumption is visible (indicator EPI), without showing a significant decrease

HDVs

Distance (km)

Avg. mass (kg)

Avg. speed (km/h)

Total fuel (l)

Avg. fuel (l/km)

Total CO2 emission (kg)

Avg. CO2 emission (g/ km)

API (kW h/ (100 km * t))

EPI (l/ (100 km * t))

Stand still (%)

Grade work (%)

Braking index (m/ s2)

Fuel variation (%)

EPI variation (%)

API variation (%)

Before After

196,656 119,457

35,225 35,225

44.60 43.72

72,514 41,630

0.369 0.348

188,535 108,238

958.71 906.08

2.06 1.98

1.05 0.96

0.11 0.13

0.22 0.21


0.22
0.19

–
5.5%

–
8.8%

–
3.7%

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

Table 3 HDV performances registered with LCMM method.

105

106

LCVs

Distance (km)

Avg. mass (kg)

Avg. speed (km/h)

Total fuel (l)

Avg. fuel (l/km)

Total CO2 emission (kg)

Avg. CO2 emission (g/k) m

API (kW h/ (100 km * t))

EPI (l/ (100 km * t))

Stand still (%)

Grade work (%)

Braking index (m/ s2)

Fuel variation (%)

EPI variation (%)

API variation (%)

Before After

64,799 58,118

15,222 15,222

41.93 41.06

14,997 13,345

0.231 0.230

38,993 34,698

601.76 597.03

1.89 1.68

1.37 1.30

0.20 0.23

0.10 0.08


0.27
0.20

–
0.8%

–
4.8%

–
10.9%

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

Table 4 LCV performances registered with LCMM method.

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

107

of the average speed. This positive result can be explained by considering the improvement of the driving style, guaranteed by a more gradual acceleration and soft hard braking (see indicators API and breaking index). 4.3. Discussion of the results The reduction of fuel consumption granted by the eco-driving field trial in Jiangsu presents values that are partially in line with most of the literature presented in Section 2. If we refer to HDVs (
5.5%, Table 3), results are consistent with the European framework to evaluate the efficiency of the ITS related to specific measures (Antonissen et al., 2013), whose average reduction is calculated in a range comprised between 5 and 10%. However, when we consider LCVs (Table 4), no significant variations before and after the training have been registered. These results should probably be better framed by considering some aspects regarding road congestion and the trips performed by the two types of vehicles. While HDV trips were mostly made on extra-urban roads, the LCV ones occurred mostly in urban areas.2 Here, many acceleration and braking events occur, as well as high standstill fuel losses. The phase 2 (after the eco-drive trainings) was executed within December and January, when road transport in China is dominated from domestic preparations for Chinese New Year holiday, the most important national holiday. Coherently, in this period higher urban congestion indexes are revealed. This condition affects mostly LCVs, and it could contribute to explaining the increase of the indexes regarding the standstill phase. These results suggest that the potentialities of eco-driving trainings could possibly generate higher results in other periods of the year. On the other hand, the economic reward given to drivers could have determined an overestimation of the real potentialities deriving from the adoption of this measure: indeed, the stimulus of drivers to a more virtuous driving style is likely to decrease if the economic incentive is not increased. Similarly, Beusen et al. (2009) demonstrated that people have the tendency to forget eco-drive training in the long term: during some monitoring activities, most drivers showed an immediate improvement in fuel consumption in response to eco-drive training, but some drivers tended to fall back to original driving habits. If the eco-drive approach proposed by the Linsen group is meant as a long-term strategy to reduce their fuel consumption and CO2 emissions, then adequate periodical updates of the trainings are required and should be planned from the beginning as an integrative part of the measure. In this long-term strategy, the LCMM approach could represent an appropriate technical solution to monitor the performances towards the achievement of a reduction of carbon emissions caused by freight transport, focussing not only on the aggregate fuel consumption value, but also on the single components that concur to define it. However, transport companies must be aware that these evaluations require adequate computational efforts, which can be addressed only through efficient big data management. A trip of one hour already generates an array of 3600 data sets. A vast amount of Giga-Byte data was collected during the whole Jiangsu field trial. This amount of data is expected to increase rapidly for larger fleets using GPS equipped smart-phones regularly. This computational aspect raises another methodological issue, which needs a further caveat. This section has presented the results regarding API and EPI before and after the field trial, in accordance with the request of the Linsen Logistics group. A more comprehensive evaluation based on a difference-in-differences analysis could have provided more robust results. To this aim, the analysis of the performances obtained by a control group who did not undergo the training could have contributed to a more complete understanding of the eco-driving effectiveness. This would have required additional data and costs, due to the technological equipment on HDVs and LCVs, which were not affordable in the context of this project. Nevertheless, future researches should include this methodological aspect into their evaluation, in order to obtain more exhaustive results. Despite this aspect, the sample of 25 vehicles (15 HGVs and 10 LCVs) considered in this field trial is vast, especially if compared to similar studies (see Section 2) and it can contribute to the scientific debate about the real impact of eco-driving on carbon emissions. Finally, a mention about the boundaries of the system has to be provided, related to the type of analysis to be performed. The methodology presented in this article has quantified the Tank-To-Wheel phase, since the aim was to evaluate the impact of eco driving on fuel consumption, thus considering neither how the fuel has been produced, nor the role of less polluting means of transport. However, in a previous piece of research (Nocera & Cavallaro, 2016b), we have underlined the importance for planners and policy-makers to cover the entire Well-To-Wheel process. This is not a contradictory aspect: when a more comprehensive environmental analysis has to be performed, which aims at assessing the total impact deriving from transport, eco driving can be only one of the measures to be adopted. In this case, the role of the Well-To-Tank energy consumption is not negligible. Accordingly, this phase should be included into more general environmental evaluations (Svennson, Møller-Hols, Glöckner, & Maurstad, 2007). 5. Conclusions In this paper, we have assessed the efficiency of eco-driving as a measure meant to reduce the fuel consumption of freight transport. To calculate the real impact of this measure, a specific Advanced Telematics Platform called Low Carbon Mobility

2 At aggregate level (i.e., including both HDVs and LCVs), 32.27% of total trips analysed in this section occurred under urban driving conditions, while the remaining 67.73% occurred under extra-urban conditions.

108

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

Table 5 Definition of the values that determine the star range. Star

Avg. speed (km/h)

EPI (l/(100 km * t))

API (kW h/(100 km * t))

Braking index (m/s2)

Stand still (%)

Grade work (%)

HDVs 1 2 3 4 5

620 (20,40] (40,45] (45,50] >50

>2 (1,2] (0.9,1] (0.8,0.9] 60.8

>3 (2.2,3] (2,2.2] (1.6,2] 61.6

6
2 [
0.5,
2) [
0.2,
0.5) [
0.1,
0.2) >
0.2

>0.1 (0.05,0.1] (0.02,0.05] (0.01,0.02] 60.01

>0.7 (0.45,0.7] (0.35,0.45] (0.25,0.35] 60.25

LCVs 1 2 3 4 5

620 (20,40] (40,45] (45,50] >50

>2 (1.5,2] (1.3,1.5] (1.1,1.3] 61.1

>3 (2.2,3] (1.8,2.2] (1.6,1.8] 61.6

6
1 [
1,
0.5) [
0.25,
0.5) [
0.2,
0.25) >
0.2

>0.1 (0.05,0.1] (0.03,0.05] (0.01,0.03] 60.01

>0.5 (0.2,0.5] (0.18,0.2] (0.11,0.18] 60.11

Fig. 6. Radar chart with results of eco-driving training for HDVs (left) and LCVs (right).

Management Cloud Platform Architecture has been developed and tested in the Chinese province of Jiangsu, showing a saving of fuel up to 5.5% for HDVs, but not substantial changes for LCVs. This micro-modelling approach operates at a high level of complexity. It provides more accurate results than the macro approaches, which are based only on the fuel sold, as collected from fuel bills or recorded in financial booking systems linked to fuel payment cards. Indeed, it is known from literature that this value does not reflect the fluctuations in fuel consumption happening along the number of trips in between filling up the tank. The advantage of the method presented here lies in the detailed understanding of driving behaviour (internal and psychological factors) always linked to the external conditions of the road network including altitude or traffic congestions. In this way, drivers can be informed instantaneously about their driving style and correct it. Even if the model has guaranteed good results in terms of reliability, further refinements of the system have to be provided, in order to obtain more accurate values for trips with mileages shorter than 5 km, with average speed values lower than 5 km/h and standstill above 90% of the entire trips. Currently, these trips, which are a marginal part of the sample collected, present abnormal values and cannot be included into the evaluation. However, in order to guarantee full replicability and scalability of the method, this component cannot be ignored, particularly when the urban freight transport has to be assessed. The eco-driving measure has revealed its potentialities as an integrative measure to reduce fuel consumption and carbon emissions, provided that it is introduced at an adequate scale and with the necessary supporting measures. Indeed, one of the main problems of eco-driving lies in its initial difficulty to be adopted. A rigorous education program, aimed at explaining to the drivers the benefits of an eco-driving style, is a first, unavoidable step to grant its diffusion. Coherently, it is becoming a central part of several awareness campaigns regarding mobility and environment (Wengraf, 2012). Barkenbus (2010) proposed to integrate this aspect with some complementary regulatory tools: reducing and enforcing interstate speed limits, making eco-driving training widely available and free to individuals, cost sharing with national governments the costs of purchasing and installing driver feedback devices. The implementation of an eco-driving approach can also be supported by the private sector, conceiving it as an integrative measure to the ‘‘pay-as-you-drive” form of insurance. According to this scheme, a first step is that existing lump sum premiums vary in proportion to the kilometres driven. As a second step, a ‘‘payhow-you-drive’’ insurance program may be proposed, which is based not only on the distance run, but also on the style

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

109

adopted by the drivers. In this case, it is fundamental the use of a reliable specific vehicle feedback device (such as the LCMM presented in this paper), able to monitor both instantaneous and total fuel consumption and carbon emissions. The implementation can be also gradual, starting with a reduction in vehicle insurance rates for those completing an eco-driving training course and demonstrating eco-driving performance on the road. The introduction of economic incentives, even if undoubtedly helpful (at least in a first phase), may not be the best solution in the long-term, especially if adequate resources are not forecast from the beginning. Indeed, it has been demonstrated (Stillwater & Kurani, 2013) that people tend to forget their behaviours, when economic motivations are either inadequate or limited in time and in some experiments, nonmonetary benefits have been more appreciated (Schall & Mohnen, 2015). However, policy makers should be aware that eco driving is only one of the numerous measures that can be adopted to reduce the carbon emissions caused by the transport sector. The adoption of this measure must be included in a broader framework: on the one hand, it should be based on a comparison of the carbon efficiency with other measures. On the other hand, the evaluation should include a more comprehensive overview of other transport impacts, such as the reduction of traffic congestion, or the improvement of alternative transport modes that the measure itself can grant. Currently, the numerous scientific and economic uncertainties (discussed thoroughly in Nocera, Tonin, & Cavallaro, 2015a) make GHG reduction only an ancillary component and not as one of the main objectives of mobility plans. Consequently, CO2 emissions assume a secondary role in the definition of the measures to be introduced. In Nocera, Tonin, and Cavallaro (2015b), we have identified the Sustainable Urban Mobility Plans (Wefering, Rupprecht, Bührmann, & Böhler-Baedeker, 2013) as the privileged form of transport plan able to solve this open issue. This remains a fundamental responsibility for policy-makers, which they should take with the support of adequate evaluation tools: the methodology presented in this paper can contribute to this ambitious but unavoidable aim.

References Alam, M. S., & McNabola, A. (2014). A critical review and assessment of eco-driving policy & technology: Benefits & limitations. Transport Policy, 35(2014), 42–49. Antonissen, T., Dijkstra, A., Dreher, S., De Hann, R., Heiber, I., Van der Kroon, P., . . . Johansson, H., (2013) Identifying the most promising ITS solutions for clean and efficient mobility, Working Group for Clean and Efficient Mobility (WG4CEM), Mobility Forum. [05.02.2016]. Baric´, D., Zovak, G., & Periša, M. (2013). Effects of eco-drive education on the reduction of fuel consumption and CO2 emissions. Promet – Traffic & Transportation, 25(3), 265–272. Barkenbus, J. N. (2010). Eco-driving: An overlooked climate change initiative. Energy Policy, 38(2010), 762–769. Beusen, B., Broekx, S., Denys, T., Beckx, C., Degraeuwe, B., Gijsbers, M., ... Panis, L. I. (2009). Using on-board logging devices to study the longer-term impact of an eco-driving course. Transportation Research Part D: Transport and Environment, 14, 514–520. Black, W. R. (2010). Sustainable transportation: Problems and solutions. New York: The Guilford Press. Boriboonsomsin, K., Vu, A., Barth, M., (2010). Co Eco-Driving: Pilot Evaluation of Driving Behavior Changes among U.S. Drivers, University of California Transportation Center UCTC-FR-2010-20. [01.04.2016]. Cavallaro, F., Maino, F., & Morelli, V. (2013). A new method for forecasting CO2 operation emissions along an infrastructure corridor. European Transport – Trasporti Europei, 55, 1–25. Demir, E., Bektasß, T., & Lavorate, G. (2011). A comparative analysis of several vehicle emission models for road freight transportation. Transportation Research Part D, 16, 347–357. DGET, Directorate-General for Energy and Transport, (2006). Mid-term review of the European Commission’s 2001 White Paper on Transport. [01-02-2011]. Dray, L. M., Schäfer, A., & Ben-Akiva, M. E. (2012). Technology limits for reducing EU transport sector CO2 emissions. Environmental Science & Technology, 46 (2012), 4734–4741. EC, European Commission, (2011). White paper: Roadmap to a Single European Transport Area – Towards a competitive and resource efficient transport system. [26.07.2015]. EC, European Commission, (2015). 2030 framework for climate and energy policies. [26.02.2015]. EU, European Union, (2014). EU energy in figures, Statistical Pocketbook 2014. Luxembourg: Publications Office of the European Union. Exchangerates, (2013). Chinese Yuan Exchange Rates for 31/12/2013. [01-12-2016]. Jeffreys, I., Graves, G., & Roth, M. (2016). Evaluation of eco-driving training for vehicle fuel use and emission reduction: A case study in Australia. Research Part D: Transport and Environment. http://dx.doi.org/10.1016/j.trd.2015.12.017. McKinnon, A. C., & Piecyk, M. I. (2009). Measurement of CO2 emissions from road freight transport: A review of UK experience. Energy Policy, 37(10), 3733–3742. Nocera, S., & Cavallaro, F. (2014). The ancillary role of CO2 reduction in urban transport plans. Transportation Research Procedia, 3(2014), 760–769. http://dx. doi.org/10.1016/j.trpro.2014.10.055. Nocera, S., & Cavallaro, F. (2016a). The competitiveness of alternative transport fuels for CO2 emissions. Transport Policy, 50, 1–14. http://dx.doi.org/10.1016/ j.tranpol.2016.05.013. Nocera, S., & Cavallaro, F. (2016b). Economic valuation of well-to-wheel CO2 emissions from freight transport along the main transalpine corridors. Transportation Research Part D: Transport and Environment, 47(2016), 222–236. http://dx.doi.org/10.1016/j.trd.2016.06.004. Nocera, S., Tonin, S., & Cavallaro, F. (2015a). The economic impact of greenhouse gas abatement through a meta-analysis: Valuation, consequences and implications in terms of transport policy. Transport Policy, 37, 31–43. http://dx.doi.org/10.1016/j.tranpol.2014.10.004. Nocera, S., Tonin, S., & Cavallaro, F. (2015b). Carbon estimation and urban mobility plans: Opportunities in a context of austerity. Research in Transportation Economics. http://dx.doi.org/10.1016/j.retrec.2015.07.009. Qian, G., Chung, E., & Horiguchi, R. (2013). Assessing eco-driving performance of a traffic platoon at urban intersections: An experimental study. In Optimum 2013–international symposium on recent advances in transport modelling. . Robert Bosch GmbH (2014). Kraftfahrtechnisches Taschenbuch. Kraftfahrtechnisches Taschenbuch. Springer Vieweg. Rolim, C. C., Baptista, P. C., Duarte, G. O., & Farias, T. L. (2014). Impacts of on-board devices and training on light duty vehicle driving behavior. Procedia-Social and Behavioral Sciences, 111, 711–720. Samaras, Z., Ntziachristos, L., Burzio, G., Toffolo, S., Tatschl, R., Mertz, J., & Monzon, A. (2012). Development of a methodology and tool to evaluate the impact of ICT measures on road transport emissions. Procedia-Social and Behavioral Sciences, 48, 3418–3427. Schall, D. L., & Mohnen, A. (2015). Incentives for energy-efficient behavior at the workplace: A natural field experiment on eco-driving in a company fleet. Energy Procedia, 75, 2626–2634.

110

K. Ayyildiz et al. / Transportation Research Part F 46 (2017) 96–110

Sims, R., Schaeffer, R., Creutzig, F., Cruz-Núñez, X., D’Agosto, M., Dimitriu, D., ... Tiwari, G. (2014). Transport. In O. Edenhofer, R. Pichs-Madruga, Y. Sokona, E. Farahani, S. Kadner, K. Seyboth, A. Adler, I. Baum, S. Brunner, P. Eickemeier, B. Kriemann, J. Savolainen, S. Schlömer, C. von Stechow, T. Zwickel, & J. C. Minx (Eds.), Climate Change 2014: Mitigation of climate change. Contribution of working group III to the fifth assessment report of the intergovern-mental panel on climate change. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press. Sivak, M., & Schoettle, B. (2012). Eco-driving: Strategic, tactical, and operational decisions of the driver that influence vehicle fuel economy. Transport Policy, 22(2012), 96–99. Stillwater, T., & Kurani, K. S. (2013). Drivers discuss ecodriving feedback: Goal setting, framing, and anchoring motivate new behaviors. Transportation Research Part F Traffic Psychology and Behaviour, 19, 85–96. Strömberg, H. K., & Karlsson, I. C. M. (2013). Comparative effects of eco-driving initiatives aimed at urban bus drivers – Results from a field trial. Transportation Research Part D: Transport and Environment, 22, 28–33. Sullman, M. J. M., Dorn, L., & Niemi, P. (2015). Eco-driving training of professional bus drivers – Does it work? Transportation Research Part C, 58(2015), 749–759. Svennson, A. M., Møller-Hols, S., Glöckner, R., & Maurstad, O. (2007). Well-to-wheel study of passenger vehicles in the Norwegian energy system. Energy, 32 (2007), 437–445. Symmons, M. A., & Rose, G. (2009). Ecodrive training delivers substantial fuel savings for heavy vehicle drivers. In Proceedings of the 5th international driving symposium on human factors in driver assessment, training, and vehicle design, Big Sky, Montana. . UN, United Nations, (2012). Regulation No. 101: Uniform provisions concerning the approval of passenger cars powered by an internal combustion engine only, or powered by a hybrid electric power train with regard to the measurement of the emission of carbon dioxide and fuel consumption and/or the measurement of electric energy consumption and electric range, and of categories M1 and N1 vehicles powered by an electric power train only with regard to the measurement of electric energy consumption and electric range. [16.06.2016]. UN, United Nations, (2015). Paris Agreement. [18.10.2016]. Wefering, F., Rupprecht, S., Bührmann, S., Böhler-Baedeker, S., 2013. Guidelines. Developing and implementing a sustainable urban mobility plan. Brussels. [19.02.2015]. Wengraf, I., (2012). Easy on the Gas. The effectiveness of eco-driving. [01.04.2016]. Wu, Y., Zhao, X., Rong, J., (2015). The Long-Term Effectiveness of Eco-Driving Training: A Pilot Study. Proceedings of the 8th International Driving Symposium on Human Factors in Driver Assessment, Training, and Vehicle Design, (pp 212–218). [01.06.2016]. Zarkadoula, M., Zoidis, G., & Tritopoulou, E. (2007). Training urban bus drivers to promote smart driving: a note on a Greek eco-driving pilot program. Transportation Research Part D: Transport and Environment, 12, 449–451.