Real-World Comparison of Probe Vehicle Emissions and Fuel Consumption Using Diesel and 5% Biodiesel (B5) Blend
Real-World Comparison of Probe Vehicle Emissions and Fuel Consumption Using Diesel and 5% Biodiesel (B5) Blend
Real-World Comparison of Probe Vehicle Emissions and Fuel Consumption Using Diesel and 5% Biodiesel (B5) Blend
com/locate/scitotenv
Real-world comparison of probe vehicle emissions and fuel consumption using diesel and 5% biodiesel (B5) blend
Karl Ropkins a,, Robert Quinn a,1 , Joe Beebe b , Hu Li c , Basil Daham c , James Tate a , Margaret Bell a , Gordon Andrews c
b
Institute for Transport Studies, University of Leeds, Leeds, LS2 9JT, UK National Center for Vehicle Emissions Control and Safety, Colorado State University, Colorado 80523-1584, USA c Energy and Resources Research Institute, University of Leeds, Leeds, LS2 9JT, UK Received 11 September 2006; received in revised form 25 October 2006; accepted 2 November 2006 Available online 16 February 2007
Abstract An instrumented EURO I Ford Mondeo was used to perform a real-world comparison of vehicle exhaust (carbon dioxide, carbon monoxide, hydrocarbons and oxides of nitrogen) emissions and fuel consumption for diesel and 5% biodiesel in diesel blend (B5) fuels. Data were collected on multiple replicates of three standardised on-road journeys: (1) a simple urban route; (2) a combined urban/interurban route; and, (3) an urban route subject to significant traffic management. At the total journey measurement level, data collected here indicate that replacing diesel with a B5 substitute could result in significant increases in both NOx emissions (813%) and fuel consumption (78%). However, statistical analysis of probe vehicle data demonstrated the limitations of comparisons based on such total journey measurements, i.e., methods analogous to those used in conventional dynamometer/drive cycle fuel comparison studies. Here, methods based on the comparison of speed/acceleration emissions and fuel consumption maps are presented. Significant variations across the speed/acceleration surface indicated that direct emission and fuel consumption impacts were highly dependent on the journey/ drive cycle employed. The emission and fuel consumption maps were used both as descriptive tools to characterise impacts and predictive tools to estimate journey-specific emission and fuel consumption effects. 2006 Elsevier B.V. All rights reserved.
Keywords: Real-world emissions; Vehicle exhaust emissions; Diesel; Biodiesel; Regulated gaseous emissions; Fuel consumption
1. Introduction Biodiesels, diesel substitutes made from biological material, e.g. plant or animal fats, and diesel blends produced using these (commonly abbreviated Bn, where n is the volume/volume percentage biodiesel in diesel), are widely considered to be more sustainable and environmentally-friendly than conventional diesels.
Corresponding author. Tel.: +44 113 343 1785; fax: +44 113 343 5334. E-mail address: K.Ropkins@its.leeds.ac.uk (K. Ropkins). 1 Current address: Environmental Team, Mayer Brown Ltd, Lion House, Oriental Road, Woking, Surrey, GU22 8AP, UK. 0048-9697/$ - see front matter 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.scitotenv.2006.11.021
This is because the bio-component of these fuels is renewable, biodegradable, most likely less toxic than diesel and perhaps, most importantly, are not fossil-fuel derived. Historically, the use of neat biodiesels, particularly unrefined vegetable oils, and bio-rich biodiesel ( B40) blends has been associated with reduced vehicle/engine performance, e.g. impeded fuel flows, blocked fuel filters and fouled injectors and piston chambers (Bagby, 1987; Knothe et al., 1992; Graboski and McCormick, 1998 and references therein; Srivastava and Prasad, 2000). Consequently, commercial fuel producers have developed various bio-oil
268
derivatives, most commonly catalytic cracking or esterification products, that either neat or as diesel blends provide fuel characteristics and vehicle/engine performances more directly comparable with conventional diesels (Ma and Hanna, 1999; Altn et al., 2001; Demirba, 2003; Ramadhas et al., 2004). The use of such derivative biodiesels is now actively encouraged in many countries through awareness and taxation exemption schemes, e.g. the National Biodiesel Board and Alternative Motor Fuels Act 1988 in the US, and the ALTENER programme in the EU. Published research data on the exhaust emissions and fuel consumption impacts of biodiesels are somewhat contradictory, perhaps a consequence of the range of vehicle/engine drive cycles, engine modification practices and fuel types and blends used in such studies. However, some observations can be made regarding general (mass-based) trends, e.g. for the gaseous species considered here and fuel consumption: (1) Many researchers report no effect on carbon dioxide (CO2) emissions (see e.g. Ali et al., 1995; Peterson et al., 1996; Carraretto et al., 2004) but both increases (e.g.+ 5% for B30 Soya bean methyl ester biodiesel; Schumacher et al., 1996) and decreases (e.g. 15% for B100 Cotton seed oil methyl ester biodiesel; Altn et al., 2001) have also been observed. (2) Both carbon monoxide (CO) and hydrocarbons (HC) emissions most likely decreases as in almost all reported studies, regardless of biodiesel: diesel blend or bio-component source, emissions are lower for biodiesel compared to diesel. In addition, there may also be a general decrease in CO and HC emissions with increasing percentage bio-component, although the scatter of reported measurements associated with this trend is large (e.g. CO 3% to 85% and HC + 25% to 39% for B20s, alone; Graboski and McCormick, 1998 and references therein; Kalligeros et al., 2003; Raheman and Phadatare, 2004). (3) Oxides of nitrogen (NOx) emission trends are less obvious with published reports suggesting both increases (e.g. Schumacher et al., 1996; Monyem and Van Gerpen, 2001; Carraretto et al., 2004) and decreases (e.g. Peterson et al., 1996; Raheman and Phadatare, 2004). (4) Fuel consumption most likely increases slightly although the effect is small to negligible, especially for derivative-based biodiesels (e.g. 0 to 5% for B20 products; Graboski and McCormick, 1998 and references therein; TurrioBaldassarri et al., 2004). To date almost all published data on the emissions and fuel consumption impacts of biodiesels have been obtained using laboratory-based methods, most commonly engine dynamometer tests, in which an engine mounted directly on a test rig is fitted with a dynamometer, and chassis dynamometer tests, in
which a vehicle is driven on a rolling road providing dynamometer style feedbacks. Dynamometer methods provide a highly controllable environment that can be used to simulate transient driving activities, e.g., different combinations of idling, accelerating, cruising and decelerating actions similar to those undertaken in on-road vehicle journeys. The standard journeys, termed drive or test cycles, used in these methods provide highly reproducible baselines for comparative testing, e.g. the New European Drive Cycle (NEDC; EEC Directive 90/ C81/01 Emission Test Cycles for the certification of light duty vehicles in Europe) or US Federal Test Procedure-75 (FTP75; Federal Register 40CFR 85/86, EPA Test Procedures for Highway and Nonroad Engines). However, such standard drive cycles are not strictly representative of vehicles under real driving conditions (Pierson et al., 1990; Black, 1991; Ross, 1994; Joumard et al., 1995, 2000; Sjdin and Lenner, 1995; Watson, 1995; Samuel et al., 2004). Therefore, the real-world emissions and fuel consumption impacts of biodiesel could be different from these estimated using standardised dynamometer/drive cycle methods. Consequently, in recent years significant research efforts have focused on the provision of more meaningful measures of real-world vehicle emissions. The larger part of this work has been in the development of real-world dynamometer drive cycles more directly representative of the broad spectrum of local real-world driving, an approach perhaps best exemplified by the Artemis driving cycle developed within the European Commission Artemis project (see e.g. Andr, 2004). However, complementary to this work a number of methods have been developed to directly measure real-world vehicle emissions, using either in-situ approaches such as roadside remote sensing (e.g. Sjdin and Lenner, 1995; Bishop et al., 1996; Jimenez et al., 2000; Chan et al., 2004), road tunnel (e.g. Pierson et al., 1996; Kristensson et al., 2004) and inverse dispersion (e.g. Ketzel et al., 2003) methods, or in-vehicle approaches such as probe vehicle (such as Schurmann and Staab, 1990; Sjdin and Lenner, 1995; Takada et al., 2002; Daham et al., 2005; Li et al., 2005) and car-chaser (e.g. Bukowiecki et al., 2002; Gouriou et al., 2004; Kittelson et al., 2004) methods. Clearly, these real-world emission measurement methods all have inherent limitations; in-situ methods can readily monitor multiple vehicles but only at a given location (e.g. a section of road or road tunnel) and are therefore only strictly representative of local road traffic, while in-vehicle methods can readily monitor whole journeys but only for a limited number of vehicles and are therefore only strictly representative of the vehicle(s) and journey(s) studied. However, the true
269
scientific value of these methods is in their use as research tools to improve understanding of the underlying mechanisms for vehicle emissions, and to develop, validate and determine uncertainties associated with real-world drive cycles and vehicle emission inventories and models. Here, regulated gas-phase pollutant (CO2, CO, HC, NOx) emissions and fuel consumption data are presented from a probe vehicle comparative study of a diesel and a B5 biodiesel produced using that diesel. Historically, the majority of biodiesel research has focused on richer blends, e.g., B50100, because these would provide greater fossil fuel savings, and B2040, as many consider these optimum vehicle performance/exhaust emissions blends. However, most conventional diesel vehicles require little or no modification to operate with lighter biodiesel blends ( B10) and B5 blends, in particular, meet legislative standards for diesel engine manufacturer's warranties in many countries (e.g. the British Standard BS EN 590). Consequently, such B10 biodiesels are already being identified (and in some cases adopted) as the first-stage fuels in the stepwise move towards large-scale biodiesel adoption by fuel suppliers, vehicle fleet operators and local and national environmental authorities alike. Our ability to predict the emission and fuel consumption impacts from the wide-scale introduction of such B10 biodiesel blends is limited due to the lack of directly relevant research data and this paper and on-going research by the authors and colleagues is intended to contribute to current scientific understanding of the real-world emission impacts of such alternative fuels. 2. Materials and methods 2.1. Probe vehicle and on-board instrumentation A 1995 (1.8 L turbocharged EURO I diesel) Ford Mondeo LX TD was instrumented and used for all experimental work discussed herein. The main specifications of this vehicle are summarised in Table 1. Prior to this study, the vehicle has been used in other probe vehicle studies, and the general level of maintenance is considered comparable with most other vehicles of similar ages. All experiments reported here were run without engine modification (i.e. not optimally tuned for each fuel), because B10 biodiesels are intended for use with unmodified diesel vehicles. The probe vehicle was instrumented with a Horiba On-Board Emissions Measurement System (OBS1300), comprising a Data Integration Unit (DIU), heated Non Dispersive Infra Red analyser (HNDIR), zirconia
Table 1 Main specifications of probe vehicle Specification Vehicle registration year Vehicle make and model Vehicle classification Engine 1995 Ford Mondeo XLD 418 TCI 1.8 L diesel (IDI) 1753cc; overhead camshaft; compression ignition; turbo-charged; inter-cooled Bore 82.5 mm; stroke 82.0 mm; compression ratio 21.5:1 Additional information M registration
Cylinders
4 (in-line)
66 kW at 4500 rpm Maximum torque 180 Nm at 2000 rpm Emission standard EU EURO I (diesel CO 2.72(3.16) g/km; HC + passenger vehicle) NOx 0.97(1.13) g/km; PM 0.14(0.18) g/km; bracketed values conformity of production limits Emission EGR NOx control; reduction oxidation catalytic technology converter
Maximum power
type NOx and Air/Fuel ratio (AFR) analyser (NOx-AFR), Annubar (modified Pitot style) flow meter, additional invehicle and ambient air monitoring, data logger (laptop) and software, global position system (GPS), power supply unit and dedicated batteries (Fig. 1). All exhaust sampling/monitoring was conducted via a purpose-built exhaust attachment installed after foreshortening the exhaust pipe. CO, CO2 and HC concentrations were measured using HNDIR, with sampling line and analyser cells operated at 120 2 C and 87 2 C, respectively. NOx concentrations and AFR (NOx-AFR probe), exhaust flow rate (Annubar) and exhaust temperature (thermocouple) were all measured directly in-situ in the exhaust flow. Exhaust pressure was measured by open line connection and DIU-mounted pressure sensor. Vehicle speed and engine revolutions were logged from external (vehicle) inputs. In addition, ambient conditions (local temperature, pressure and humidity) were monitored using dedicated vehicle-mounted sensors. For further details of this system, see development work of Nakamura et al. (2002). 2.2. Fuels Two fuels were used in this study: (1) An ultra low sulphur diesel (EU specification, 2005; sulphur content 50 ppm) supplied by Bayford Fuels (http://www.
270
Fig. 1. System schematic for Horiba On-Board Emissions Measurement System (OBS-1300). (N.B.: Solid arrows indicate power lines, with arrow in direction of flow; broken arrows indicate signal lines, with arrow in direction of signal.).
bayford.co.uk/homepage.asp), hereafter referred to as BD; and, (2) A biodiesel (Global) also supplied by Bayford Fuels, and understood to be a blend of 5% renewable rape seed methyl ester in BD, hereafter referred to as BB5. 2.3. On-road journeys/sampling Three journey routes were used within this study, Routes 1 to 3 as shown Fig. 2AC), and all data reported here was collected during the period 19-30/07/2004. Cold-start effects were not investigated during this study. Route 1 (Fig. 2A) was a relatively simple three (left hand turn) circuit journey (distance ca. 6 km). Traffic flows were typically moderate and three of the four corner junctions are traffic signal controlled, so replicate journeys tended to be more reproducible than most realworld journeys especially if, as here, sampling was restricted to non-rush hour periods. Route 1 was considered a best case for collecting multiple journey datasets with relatively high reproducibility. Route 2 (Fig. 2B) was a single return journey through both urban and extra urban driving environments (distance ca. 16 km). Route 2 journeys typically employ a range of different driving styles/manoeuvres both in terms of general traffic conditions (from free flow to congested and steep down-hill through to steep up-hill slopes) and individual driver activities (e.g. left- and righthand turns, first- and last-exit roundabout turns, unregulated and speed camera regulated driving). By comparison to Route 1 journeys, Route 2 journeys reflect significantly more diverse driving conditions/behaviour. Route 2 was
considered a best case scenario for collecting data broadly representative of real-world driving. Route 3 (Fig. 2C) was an urban journey through predominately signalised and/or priority junctions into and around a city centre (distance ca. 10 km). Although driving manoeuvres/styles employed are similar to those associated with Route 1, traffic flows are significantly higher than on Route 1 and traffic management was more comprehensive. Route 3 was considered representative of typical managed urban driving. Multiple datasets were collected for Route 1 (8 to 13 for each fuel type) and more limited datasets (4 to 6 for each fuel type) for Routes 2 and 3, all by the same driver. 3. Data handling 3.1. Data quality assurance/quality control (QA/QC) 3.1.1. Instrument performance In bench-tests using reference gases, the measurement accuracy of the OBS-1300 was determined to be within 2% of full scale for HNDIR CO, CO2 and HC measurements and NOx-A/F probe NOx measurements. Manufacturers reported bench-test measurement accuracies of within 1% of full scale for exhaust temperature and pressure and within 3% of full scale for Annubar flow meter pressure differential. Consistent with these reports, good agreement has previously been reported for Annubar exhaust flow rate measurements and other exhaust flow measurement techniques, e.g. smooth approach orifice (SAO) flow meter and fuel flow meter/AFR, except at vehicle idle where the Annubar overestimated flow as a
271
Fig. 2. Standard vehicle journeys used within this biodiesel study and example speed time-series.
result of flow pulsation (Nakamura et al., 2002; Daham et al., 2005). A more detailed validation of OBS-1300 invehicle performance is currently underway, and is intended to be the subject of future publication.
3.1.2. Instrument calibration The OBS-1300 was routinely calibrated during study periods, in accordance with Manufacturer's guidelines. HNDIR analyte calibrations were checked and span re-
272
calibrated (as necessary) at least once every seven days (up to once a day during periods of intense use) using a standard mixture (CO2, 16% v/v; CO, 3%, v/v; hexane, 1000 ppm in nitrogen; supplied and certified by BOC Limited) span reference and pure nitrogen (zero-grade nitrogen; BOC Limited) zero reference. HNDIR analyte and Annubar exhaust flow zero measurement drifts were checked and zero-points re-calibrated (as necessary) at the start of each test journey or at least once each hour. The NOx-AFR probe NOx calibration was checked at least once every 5000 miles, and was within manufacturer's tolerances for the periods associated with this study. Calibrations, journeys and any additional comments about instrument performance and journey time events were routinely logged and reviewed as part of standard quality assurance procedures. 3.2. Data processing Although logged simultaneously, different probe vehicle measurement parameters cannot be simply time aligned using a time stamp. Measurements are functions of the point in the monitored system at which they are measured, and the sampling method (direct/insitu or displaced sub sample transfer, etc) and measurement technology (instrument response times/ resolutions) used. Consequently, monitored parameter time series and any daughter parameters calculated from them require careful data handling to minimise trend distortions resulting from misaligned data. Here, raw data was aligned and basic emissions and fuel consumption measurements determined using the following methods. 3.2.1. Basic calculations On-road data was logged at 1 Hz (1 measurement per second) during all probe vehicle journeys. Exhaust flow rate (QEX; L/min), standardised to 20 C and 1 atm (293.15 K and 101.3 kPa), was calculated as: QEX r PEX 293:15 PPITOT k TEX 101:3 DEX 1
As all parameters used to calculate QEX were measured directly in the exhaust pipe using techniques with response times 1 s, QEX was treated as an instantaneous measurement and used as the point of reference for subsequent data alignment. (Analyte emissions (nMASS; g/s), also standardised to 20 C and 1 atm (293.15 K and 101.3 kPa), were calculated using the general formula: nMASS nttDT MWTn QEX UCF 273:15 293:15
Where DT (s) is the delay time for the measurement of n relative to associated QEX measurement; [n]t=t+DT is the concentration of analyte n measured DT seconds after QEX; MWTn is the molecular weight of n (g/mol); and UCF represents required unit conversion factors. AFR measurements were aligned using: AFR AFRttDT : 3
HNDIR measurement (CO, CO2, HC) DTs were estimated as the sum of HNDIR sample line pump flown rate and instrument response time for each analyte and were determined experimentally. Other DTs not determined experimental, e.g. NOx and AFR from the NOx-AFR probe, were obtained from the manufacturers. Fuel consumption (g/s) was calculated using CO, CO2 and HC emissions measurements and a carbon balance method (FCMASS,CB; Eq. (4)) and validated using an independent method based on one AFR measurement (FCMASS,AFR; Eq. (5)): RCB;F awtC a awtH b awtO awtC awtC aEX awtH awtC
RCB;HC;EXH HCMASS awtC awtC COMASS CO2 mwtCO mwtCO2 RCB;F
MASS
RCB;HC;EXH FCMASS;CB
Where k is a proportional coefficient relating to the exhaust diameter and determined by Annubar calibration using a SAO flow meter; PEX is the exhaust gas pressure (kPa); TEX is the exhaust gas temperature (K); PPITOT is the Annubar (Pitot) pressure difference measured directly in the exhaust pipe (kPa); and DEX is the density of exhaust gas at 20 C and 1 atm (1.20 kg/m3 for all fuels considered herein).
Where RCB,F is the carbon balance in the fuel used; awtn is the atomic weight of element n (hydrogen, carbon, oxygen, etc); and are the average atomicity ratios for H/C (1.86 for all fuels considered herein) and O/C (0.00 for BD; 0.01 for BB5) atomicity ratios of fuels used
273
herein; RCB,HC,EXH is the carbon balance of hydrocarbon in exhaust gas; and mwtn is the molecular weight of compound n (CO, CO2, etc). Inserting FCMASS,CB from Eq. (4) into Eq. (5) allows the calculation of an alternative measurement of AFR (AFRCB) subject to HNDIR DTs. By comparison, the principle AFR measurement was subject to NOx-AFR sensor DTs. Therefore, the accuracy of applied DTs was tested by confirming optimum agreement between AFR measures, using 2, (AFRCB AFR / AFR)2, or linear correlation coefficient, r(AFR, AFRCB), as goodness of fit functions and DTs as the optimisation variables. 3.2.2. Total journey measurements Distance normalised total journey emissions (nTOTAL/km; g/km) and fuel consumption (FCTOTAL/km,m calculated using either carbon balance, m = CB, or AFR, m = AFR data; g/km) were calculated using Eqs. (6) and (7), respectively:
tx X
nTOTAL=km
t 1
nMASS
x tx X
t 1
or fuel consumption measurements were assumed to be constant and estimated by fit optimisation, as follows: (1) Maps were produced using a range of speed and acceleration measurement offsets ( 10 to + 10 for each, generating 100 combinations for each case); (2) Each map was used as a model to predict the emissions or fuel consumption data series used to generate it; and, (3) The best-fit map (and therefore optimum time offsets) were identified on the basis of model performance (2 and r). Vehicle exhaust flows vary with speed and acceleration. Consequently, the above assumption of constant time offsets is not strictly valid (see e.g. Weilenmann et al., 2003). However, delay times associated with diesel car exhaust flows have been estimated to vary between 0.078 to 1.3 s (or 0.7 0.6 s). Therefore, for this application, difference mapping diesel car emissions at 1 Hz resolution, the use of constant time offsets is considered an acceptable approximation. One future intended development for this method (which will be required before it can be applied to gasoline cars) is the incorporation of a compensation step to take into account exhaust flow dynamics, possibly based on the methods of Weilenmann et al. (2003) and Ajtay et al. (2005). 4. Results and discussion
FCMASS;m x 7
FCTOTAL=km;m
4.1. Initial observations In general, all journeytime data series exhibited similar trends, regardless of journey, fuel or driver, as illustrated in Fig. 3 where CO, CO2, HC and NOx exhaust emissions during a single circuit from a Route 1 journey using BD fuel. As would be expected, emissions tend to be associated with driving events as opposed to being evenly distributed throughout the journey. For example, typically 50% of CO, CO2, HC and NOx are emitted in ca. 5%, 20%, 10% and 20% of Route 1 journey time, respectively. Fig. 4 presents data from a 200 s time-series from the same journey to demonstrate the vehicle motion/ emissions and motion/fuel consumption relationships observed at 1 Hz sampling resolution. Not surprisingly, CO2 emissions broadly associate with periods of relatively high fuel consumption, as well as the positive component of acceleration (i.e. acceleration N 0 m/s2). A more direct relationship with jerk (rate of change of acceleration; m/s3) was not apparent at this sampling resolution, although at higher sampling resolutions (e.g. 10 Hz) this may provide a more direct measure of engine loading and/or vehicle work done. Similar acceleration relationships are observed for both NOx emissions and fuel consumption, although NOx emissions tend to be more strongly associated with prolonged
Where nMASS and FCMASS are the emission and fuel consumption rates (g/s), x is the distance travelled in a given journey (km) and time t = 1 to tx is the time taken (s) to complete that journey. 3.2.3. Emission and fuel consumption maps Speed/acceleration emission and fuel consumption maps were produced by grouping 1 Hz measurements according to speed and acceleration and averaging each of these data bins to produce a point on a response surface. (Data were typically distributed unevenly across the speed/acceleration surface with the largest variation in measurement values often observed near the speed/acceleration origin. Therefore, data bin dimensions were varied across observed ranges as follows: Speed b 0.05, 0.05 b 0.5, 0.5 b 1.5, 1.5 b 2.5, 2.5b 3.5, etc., to 34.5b 35.5, 35.5 m/s; Acceleration b 4.5, 4.5b 3.5, 3.5b 2.5, 2.5b 1.5, etc., to 3.5b 4.5, 4.5 m/s2 .) Although probe vehicle movement, emissions and fuel consumption are logged simultaneously, associated events do not occur at the same time because all these engine action responses are subject to different delay times. Here, for a diesel car time offsets between speed, acceleration and emissions
274
acceleration by comparison to CO2, most likely reflecting the relatively high engine temperatures required to generate NOx by comparison to CO2. Such observations are broadly consistent with those of Takada et al. (2002) for CO2 and NOx exhaust emissions from a diesel freight vehicle during real-world journeys studied in Yokohama, Japan. HC emissions broadly associate with relatively high AFR and deceleration (i.e. acceleration b 0 m/s2), perhaps reflecting lower engine fuel consumption and exhaust catalyst conversion efficiencies during periods of high AFR. CO emission relationships are more complex with emission events associating with both the early stages of acceleration and the later stages of deceleration (perhaps gearing-down manoeuvres). 4.2. Total journey emissions and fuel consumption comparisons In conventional dynamometer/drive cycle studies a series of driving events are replicated on a test bed and emissions and fuel consumption measurements collected and reported as total journey measurement, e.g. as emissions or fuel consumed per kilometre travelled, and these measurements, in turn, commonly form the basis for most conventional dynamometer-based fuel comparison studies. Therefore, probe vehicle total journey measurements, which are broadly analogous to such conventional dynamometer/drive cycle measurements, would on first consideration appear a practical option for real-world fuel comparison studies. However, by comparison to dynamometer drive cycles, which can be rigorously defined and replicated, real-world journeys are more ad hoc in occurrence with the driver responding to a series of unique events (driving conditions, other drivers' behaviour, etc.) even on journeys where routes have been carefully defined. The larger scatter associated with probe vehicle total journey measurements (although a direct measure of real-world variations rather than an artefact of technique accuracies and therefore of undoubted research interest) presents a significant hindrance to the use of this approach in comparative studies. Here, probe vehicle data are used to demonstrate the limitations of total journey measurements in comparative studies even when rigorous standardisation is applied to the sampling procedure. Multiple Route 1 journeys were undertaken using BD and BB5 fuels, resulting in 8 and 13 complete journeys, respectively. Sub-samples of comparable total journey measurements for both experimental cases were selected by excluding atypical journeys (here, those significantly outside overall average journey times and distances, i.e., outside 11.5 2 min and/or 5.77 0.05 km, reducing
available datasets to 7 for BD and 9 for BB5, respectively). Probe vehicle total journey emission and fuel consumption measurements obtained using this approach are summarised in Table 2. Both here and elsewhere (e.g., Takada et al., 2002; Beebe et al., 2003), it has been demonstrated that individual events (e.g., periods of harsh acceleration or overtaking manoeuvres) can significantly affect total journey emissions. Therefore, this standardisation approach, which is comparable to (but admittedly cruder than) the exclusion of bad drive cycles in conventional dynamometer-based studies, provides a robust means of minimising the impact of atypical events and other similar confounders on comparative studies such as this, where vehicle performance (e.g. technology or fuel) is assessed. Table 2 demonstrates the main limitation of probe vehicle/total journey experiments by comparison to dynamometer drive cycle measurements. Even here where significant efforts have been made to standardise probe vehicle journeys, samples of replicate total journey emissions and fuel consumption measurements exhibit marked variance with relative standard deviations of between 6 and 96% depending on analyte. By comparison, conventional dynamometer drive cycle measurements are typically more reproducible, e.g. see TurioBaldassarri et al. (2004) which reported relative standard deviations of ca. 1 to 23% for the same analytes. Furthermore, the dependence of real-world probe vehicle sampling on external conditions (e.g., weather conditions, driving conditions and the behaviour of other drivers) means that data collection is often more arduous and time consuming than for laboratory-based dynamometer studies. This combination of relatively high variance and practical limitations to sampling set size significantly hinders the interpretation of such data, most notably dataset distributions (e.g. normal or log normal) and therefore most appropriate statistical difference tests cannot be conclusively identified. Here, a robust approach similar to that described by Kauman et al. (1956) for small datasets was applied: (1) Three statistical tests based on different data distribution assumptions were applied (Student t-test which assumes normal distribution, and two nonparametric tests, one exact, Fisher Pitman, and one ranked, Wilcoxon Mann Whitney). And, (2) The degree of confidence associated with the observed differences scored according to the number of tests proving positive (i.e. No for P 0.05 for all three tests; Yes for P b 0.05 for all three tests; Possible for cases in between). Unfortunately, although robust, this approach does not provide scientifically convincing results in this case. When comparing total journey measurements for BD and BB5 fuels: (1) No significant differences were
275
Fig. 3. Journeytime emissions of CO, CO2, HC and NOx during a single circuit of Leeds Hyde Park (from a Route 1 journey by 1995 1.8 L Diesel Mondeo and Driver 1 using BD fuel). Second-by-second emissions (g/s) depicted as GPS-located circle plot with diameter proportion to emissions.
observed for CO and HC emissions (3/3 tests negative), although here differences would have needed to be large to be observable given the particularly high variances
associated with these measurements (relative standard deviations of 1996%). (2) A possible significant difference of 813% was observed for NOx emissions
276
Fig. 4. Journeytime vehicle motion (speed, acceleration and jerk), exhaust emissions (CO, CO2, HC and NOx) and fuel usage (AFR and consumption) during ca. 200 s of a Route 1 journey (1995 1.8 L Diesel Mondeo; Driver 1; BD fuel). Two periods (one acceleration, one deceleration) are identified (regions enclosed by dotted lines) to highlight emission relationships.
(1/3 tests positive). And, (3) A significant difference of 78% was observed for fuel consumption (3/3 tests positive) but no difference for CO2 emissions (3/3 tests
negative), a particularly unusual observation given the high correlation between fuel consumption and emissions of CO2, the main product of fuel combustion. (N.B.
Table 2 Distance normalised total journey Route 1/Driver 1 exhaust emissions (CO, CO2, HC and NOx; g/km) and fuel consumption (g/km) BD, 1n = 7 Mean CO CO2 HC NOx Fuel consumption 0.121 216.8 0.076 0.472 68.54 Median 0.073 220.9 0.077 0.461 69.89
2
BB5, 1n = 9 Standard deviation Mean 0.148 231.6 0.073 0.511 74.06 Median 0.122 233.3 0.064 0.523 74.55
2
Standard deviation
0.116 (96%) 14.13 (7%) 0.014 (19%) 0.029 (6%) 4.490 (7%)
0.129 (87%) 15.83 (7%) 0.023 (32%) 0.040 (8%) 5.036 (7%)
No (P = 0.666; 0.650; 0.634) No (P = 0.073; 0.075; 0.064) No (P = 0.814; 0.805; 0.315) Possible (P = 0.047; 0.052; 0.064) + 8 (mean), +13% (median) Yes (P = 0.039; 0.044; 0.030) + 8% (mean), +7% (median)
1 Route 1 selection criteria: 11.5 2 min journey time, 5.77 0.05 km journey distance (7/8 and 9/13 for BD and BB5, respectively). 2 Standard deviation reported with relative standard deviation (standard deviation / mean 100%) in parentheses. 3 Statistical significance of difference reported in form P = Student; Fisher Pitman; Wilcoxon Mann Whitney at 95% confidence level (P 0.050). Significance assignment: All three tests score non-significant, No; All three tests score significant, Yes; Else, Possible.
277
Sensitivity testing indicated that the different fuel parameters used for BB5 and BD fuels in Eq. (4) contributed to, but did not completely explain fuel consumption trends.) The statistical strength of these probe vehicle total journey measurements were further investigated by estimating the standard error trends using a re-sampling method, as follows: Measurements within a given set (e.g. CO emissions for BB5) were randomly sampled without replacement to produce daughter datasets with sample sizes (n) over the range 3 to n. The process was repeated 1000 times for each set and standard errors calculated. The results of this procedure are presented as confidence interval plots in Fig. 5. Although in each case standard error/confidence interval trends tend towards uniformity, uniformity is clearly not achieve in any case, indicating that average measurements are probably not strictly
statistically representative. Furthermore, the extrapolation of these trends indicates that significantly larger datasets of the order n = 25 to 75 would most likely be required to achieve unambiguous statistical representation. However, even with a larger dataset it is unclear whether the statistical rigour of the experiment would improve because over the longer sampling period necessary for a larger study, the variance associated with study confounders, such as driver behaviour, local driving conditions and ambient conditions, could also increase. 4.3. Emissions and fuel consumption map comparisons If the probe vehicle data could be analysed at its 1 Hz sampling resolution this would provide a much larger dataset and potentially greater statistical confidence.
Fig. 5. Standard error trends (mean 95% confidence interval; 1.96 standard error) as a function of sampling size (n) for total journey exhaust emissions (CO, CO2, HC and NOx) and fuel consumption during Route 1 journeys (1995 1.8 L Diesel Mondeo) using BD and BB5 fuels.
278
Fig. 6. CO, CO2, HC and NOx emission and fuel consumption speed/acceleration maps for probe vehicle (1995 1.8 L Diesel Mondeo) using BB5 and BD fuels. Top to bottom: BB5 map (left scale), BD map (left scale), direct difference BB5-BD (right scale), difference after exclusion of unpaired and non-statistically different datasets (right scale). All scales are presented in g/s. All maps drawn using Surfer Version 8.05 (Golden Software, 2002), applying Natural Neighbour data handling.
279
Fig. 6 (continued ).
However, the en masse comparison of two 1 Hz datasets collected in multiple-manoeuvre journeys, regardless of sample size, is subject to the confounding influence of atypical in-journey events. Here, we propose that it is more logical, data-efficient and (as demonstrated below) informative to sort 1 Hz datasets according to associated driving activity and compare sub-samples of data collected under similar driving conditions. Here, surface maps have been selected as the basis for difference testing because these can be readily applied to a wide range of data types regardless of the underlying mechanisms associated with emissions or fuel consumption. Vehicle speed and acceleration were used as the mapping parameters because the resulting maps are more directly related to driving activity/ journey type than other commonly used mapping
parameter combinations, e.g. vehicle speed / (speed acceleration) or engine speed/torque. Difference maps were generated using the following method: (1) Speed/acceleration emission and fuel consumption maps were produced for compared cases as described in Section 3.2. (2) Preliminary difference maps were produced by subtracting the maps for one case from those for the comparison case. (3) Difference map speed/ acceleration points were excluded if one or both parent maps contained no associated measurements (i.e., an unmeasured difference rather than no difference). And, (4) Each different map speed/acceleration point was validated using a Wilcoxon Mann Whitney difference test and points were excluded if found not to be statistically significant at the 95% confidence level (i.e., not different and reset to zero) or insufficiently populated for testing (i.e.,b 6 entries
280
in either or both parent maps). (N.B.: The distributions of sub-sampled data populations typically varied across parent speed/acceleration map surfaces and occasionally contained likely outlier points, i.e., 12 points significantly outside the range of other points grouped with it/them. Therefore, Wilcoxon Mann Whitney was selected as the most appropriate/robust difference test for use with this method because it is nonparametric, i.e., it makes no explicit assumptions about data distributions unlike most conventional difference tests, and ranked, i.e., it is less subject to outlier effects than absolute methods.) Fig. 6 presents emission and fuel consumption difference maps for Route 2 journeys using BB5 and BD fuels. The method provides two levels of information: (1) Final difference maps (those produced after the exclusion of unpaired and statistically non-different data) represent the differences between compared cases
that can be most confidently attributed to the compared difference (fuel type), i.e., the statistically significant differences observed over similar speed/acceleration ranges. And, (2) The differences between the initial and final difference maps represent the indirect differences between compared cases, i.e., those differences that are most likely the result of different speed/acceleration distributions in the two sample sets (e.g. resulting from different fuel performance, driver behaviour or driving conditions). CO2 and NOx emissions and fuel consumption data produce classic speed/acceleration map distributions with the highest measurements in the high speed/high acceleration section of the speed/acceleration surface, consistent with acceleration and speed relationships observed during the initial inspection of the data. HC emissions data produce reversed speed/acceleration
Fig. 7. Route 2 speed/acceleration map predictions of NOx emissions of probe vehicle (1995 1.8 L Diesel Mondeo) using BB5 and BD fuels during different vehicle journeys and drive cycle. Top: Comparison of actual probe vehicle measurements and model predictions for Route 1 and 3 Journeys. Bottom: Comparison of BD and BB5 NOx emissions for NEDC and FTP75 drive cycles and their component stages/phases. (N.B.: FTP75 phase 1 excluded because model does not account for cold start emissions.).
281
maps with high emissions most strongly associated with deceleration, again consistent with previous observations (see Section 4.1). CO emissions data produces relatively usual speed/acceleration maps with high emissions most strongly associated with low speed/ high acceleration events. Many of the highest measurements in the BD CO2, NOx and fuel consumption maps are observed at speeds greater than 25 m/s and/or at accelerations above 2 m/s2, a region where no data were collected on any journey conducted using BB5 fuel. Although about 23% of BD data was outside the range observed with BB5, these relatively high measurements typically make a relatively large contribution to associated BD total journey measurements (ca. 5% each for CO2, NOx and fuel consumption). An opposite but less distinct trend is observed with BD CO and HC maps because the high speed (N 25 m/s) regions of these maps are typically associated with low-to-medium emissions (ca. 1 2% of CO and HC total journey emissions). These indirect differences account for most of the differences estimated for total journey measurements. Although such indirect differences could result from differences in driver behaviour/driving conditions, it is worth noting that these differences were consistent (i.e., all BD data series contained data for speeds above 25 m/s; no BB5 data series contained data for speeds above 25 m/s). Similarly, speeds and accelerations recorded during a series of runs produced using a garage forecourt diesel prior to the study were highly similar to those for BD data. Therefore, these observations suggest B5 fuels may have an affect, albeit relatively small, on vehicle performance. In other words a vehicle using B5 fuels may not be as responsive as those using conventional diesels, although more comprehensive testing, most likely a combination of probe vehicle and dynamometer studies, would be required to confirm these preliminary observations. By comparison, direct differences between emission and fuel consumption maps for BB5 and BD tend to be less pronounced; in general, small positive and negative differences are observed at different areas on the speed/ acceleration surface. This indicates that in situations where driving speed/acceleration remain similar the (direct) effect of substituting a conventional diesel with a B5 fuel is small and highly dependent on the distribution of speed/acceleration events undertaken (i.e., the drive cycle run or the journey undertaken). 5. Further discussion One important consideration in any research project or programme is how to best use different equipment and/or approaches.
5.1. Probe vehicle and dynamometer research Section 4.2 demonstrates the main (and unavoidable) limitation of probe vehicle/total journey comparative studies by comparison to conventional dynamometer/ drive cycle comparative studies: Probe vehicle/total journey measurements are simply not as reproducible as conventional dynamometer/drive cycle measurements. Various strategies could be applied to improve the reproducibility of probe vehicle/total journey data, e.g. driver behaviour standardisation, journey selection criteria and/or increased data sampling volume. However, these strategies may only be of limited value because selection criteria reduce the real-world value of the data, whilst larger sampling volumes may even increase (rather than reduce) measurement variability because variability associated with the external confounders (driving conditions, etc) may be larger during the longer study periods needed to produce larger datasets. Therefore, it is unrealistic to expect to be able to produce real-world probe vehicle/total data that is as statistically convincing as conventional (laboratory) data derived from dynamometer/drive cycle studies. Consequently, if standard emissions measurements are required (e.g. for emissions legislation, engine manufacture guidelines or even technology development), a dynamometer/drive cycle approach is clearly more appropriate than a probe vehicle/total journey approach. However, the principle value of probe vehicle research is in its potential to produce representative real-world data, i.e. to study the real-world impact of confounders such as driver behaviour, driving conditions and ambient air quality that dynamometer-based studies cannot readily address. So, while dynamometer/ real-world drive cycles most likely provide the best (and least unambiguous) means of making standard emissions measurements (e.g. emissions standard adherence), probe vehicle studies provide a practical means of quantifying the error bars associated with such standards measurements when extrapolated to the realworld. As a result probe vehicle research provides valuable feedback for the validation and fine-tuning of dynamometer/real-world drive cycles, emissions inventories and models, vehicle technology and fuel development. Here, for example, final difference emissions and fuel consumption maps demonstrate that the direct emission and fuel consumption differences observed for diesel and B5 substitutes are likely to be very small. However, the indirect differences, which are, to some degree, a function of different vehicle performance/ driver behaviour patterns and indicated by the difference between initial and final difference maps, are, in the
282
journey-set considered here at least, by comparison much larger. Conventional dynamometer-based drive cycles will, by their nature, measure predominantly direct (same speed/acceleration) differences. Although the different map methods presented here are still in development and may be subject to further refinements, these initial results suggest that an important additional consideration could be that a vehicle fleet change may have an indirect effect (e.g., the introduction of a new fuel affecting vehicle performance and thereby driving behaviour) that would not be quantified by dynamometer studies alone and which could be potentially and significantly larger than the more readily measured/ quantified direct effect. 5.2. Difference map methods The speed/acceleration emission and fuel consumption difference mapping methods presented here represent an initial step in developing a meaningful, statistically robust and data-efficient means of analysing and reporting results that could in future be applied in both probe vehicle and dynamometer studies. Although these methods and the biodiesel work described here are part of on-going research, and as such are likely to be subject to refinements in future studies, the mapping approach provides a more flexible reporting format than conventional dynamometer/drive cycle (or probe vehicle/total journey) measurements. In a conventional dynamometer/drive cycle comparison study, results would typically be reported as percentage changes in emissions or fuel consumption based over the whole drive cycle. This measurement of change approach can then only be readily extrapolated to other similar driving situations, i.e. real-world cases where driving conditions/driver behaviour is similar to that in the measured drive cycle. If the real-world case incorporates significant off-cycle driving or even a significantly different distribution of driving events, the uncertainty associated with the extrapolation of that change measurement to the local vehicle fleet increases markedly. By contrast, the difference map provides a simple format for describing the emission and fuel consumption impacts of a given change over a range of different journey types. Therefore, once refined and validated, this approach could provide a directly useful output for potential endusers, e.g. local authorities, traffic managers and emissions abatement technology developers. At the simplest level it can be used qualitatively to make general observations regarding emission and fuel consumption trends. For example, the NOx difference maps indicate that traffic networks where congestion
(and therefore low speed stop/start driving) is more commonplace are likely to be subject to more significant increases in NOx emissions upon the introduction of B5 fuels by comparison to more free flowing areas. However, bothparent maps and daughter difference maps could also be used quantitatively as models to predict emissions and/or fuel consumption during different journeys using the speed time-series as an input, if it can be appropriately validated. To demonstrate this approach Route 2 speed/acceleration NOx emission maps for BD and BB5 were used to predict NOx emissions for two drive cycles (NEDC and FTP75 phase 2 and 3) and the results presented in Fig. 7. Probe vehicle measurements for Route 1 and Route 3 journeys and the Route 2 NOx emission map predictions of these are presented alongside as a measure of the performance of this approach. In general, although the emission map models tend to underestimate emissions (ca. 97% for Route 1; ca. 90% for Route 3), predicted total emission trends are similar to actual measurements, i.e., model predictions of 6.9% and 4.4% NOx emission increases for the probe vehicle using BB5 by comparison to BD for Routes 1 and 3, respectively, and observed increases of 8.7% and 5.3%. Predicted NOx emission trends for the standard drive cycles compared here range for + 4.8% (for the NEDC extra urban drive cycle component, EUDC) to + 21% (for the NEDC urban drive cycle component, UDC), indicating that driving activity/behaviour could have a significant effect on the real-world impact of the introduction of biodiesels, even low bio-content blends such as the B5 fuel studied here. 6. Conclusions This study demonstrates that probe vehicles provide useful research tools for the investigation of real-world vehicle emissions and fuel consumption impacts of fuels. However, given the inherent variability in realworld measurement, great care needs to be taken when analysing and interpreting the data. At the total journey measurement level, data presented here indicates that replacing diesel with a B5 substitute could result in significant increase in both NOx emissions (813%) and fuel consumption (78%). By comparison, no significant effect was observed for CO, CO2 and HC, although in the cases of CO and HC sampling variance was large and this may have obscured small trends, e.g. changes of a few percent. However, statistical analysis clearly demonstrated the limitations of total journey measurement approaches when handling real-world data. At 1 Hz resolution, speed/acceleration difference map methods, developed as part of this study, indicate
283
that differences in the speed/acceleration distribution of the journeys may account for much of the observed variation. This would appear to indicate that interjourney rather than inter-fuel differences had the largest impact on emissions and fuel consumption, although at this stage it should be emphasised that it is unclear to what degree the fuel change impacted on vehicle performance and thereby driver behaviour. By comparison, direct (statistically significant) differences between emission and fuel consumption maps for BD and BB5 fuels tend to be less pronounced, although as small positive and negative differences are observed at different areas on the speed/acceleration surface, this indicates that the effect is highly dependent on the distribution of speed/acceleration events undertaken (i.e., the drive cycle run or the journey undertaken). The difference map method provides a flexible output for third party emissions and fuel consumption impact estimation and therefore could provide a useful format for reporting both probe vehicle and conventional dynamometer study results. Acknowledgements The authors would like to thank the UK EPSRC for a JIF (Joint Infrastructure Fund) award (research grant GR/M88167/01) for the LANTERN project (Leeds Health, Air Quality, Noise, Traffic, Emissions Research Network) and the RETEMM (Real-world Traffic Emissions Measurement and Modelling) project award (research grant GR/S31136/01) and associated support of this work. The authors would also like to thank Bob Boreham and Ferri Ghahri-Saremi for significant input and advice, both technical and practical. References
Ajtay D, Weilenmann M, Soltic P. Towards accurate instantaneous emission models. Atmos Environ 2005;39:24439. Ali Y, Hanna MA, Borg JE. Optimization of diesel, methyl tallowate and ethanol blend for reducing emissions from diesel engine. Bioresour Technol 1995;52:23743. Altn R, etinkaya S, Ycesu HS. The potential of using vegetable oil fuels as fuel for diesel engines. Energy Convers Manag 2001;42: 52938. Andr M. The ARTEMIS European driving cycles for measuring car pollutant emissions. Sci Total Environ 2004;334/335:7384. Bagby MO. Vegetable oils for diesel fuel: opportunities for development. SAE Technical Paper Series. St Joseph, MI: Society of Automotive Engineers, Inc.; 1987. p. 871588. Beebe J, Bell MC, Tate JE. Analysis of modal emissions implications for air quality modelling and management. Paper presented at the University Air Quality Conference, Prague; 2003. Black FM. Control of motor vehicle emissions the U.S. experience. Crit Rev Environ Control 1991;21:373410.
Bishop GA, Stedman DH, Ashbaugh LL. Motor vehicle emissions variability. J Air Waste Manage Assoc 1996;46:66775. Bukowiecki N, Dommen J, Prevot ASH, Richter R, Weingartner E, Baltensperger U. A mobile pollutant measurement laboratory measuring gas phase and aerosol ambient concentrations with high spatial and temporal resolution. Atmos Environ 2002;36:556979. Chan TL, Ning Z, Leung CW, Cheung CS, Hung WT, Dong G. Onroad remote sensing of petrol vehicle emissions measurement and emission factors estimation in Hong Kong. Atmos Environ 2004;38(4):205566. Carraretto C, Macor A, Mirandola A, Stoppato A, Tonon S. Biodiesel as alternative fuel: experimental analysis and energetic evaluations. Energy 2004;29:2195211. Daham B, Andrews GE, Li H, Ballesteros R, Bell MC, Tate JE, Ropkins K. Application of a Portable FTIR for Measuring On-Road Emissions. SAE Technical Paper Series; 2005. Paper no 2005-01-0676. Demirba A. Biodiesel fuels from vegetable oils via catalytic and noncatalytic supercritical alcohol transesterifications and other methods: a survey. Energy Convers Manag 2003;44(13):2093109. Golden Software, Inc. Surfer Version 8.05 User's Manual, 2002. Gouriou F, Morin J-P, Weill M-E. On-road measurements of particle number concentrations and size distributions in urban and tunnel environments. Atmos Environ 2004;38:283140. Graboski MS, McCormick RL. Combustion of fat and vegetable oil derived fuels in diesel engines. Pror Energy Combust Sci 1998;24:12564. Jimenez JL, McManus JB, Shorter JH, Nelson DD, Zahniser MS, Koplow M, et al. Cross road and mobile tunable infrared laser measurements of nitrous oxide emissions from motor vehicles. Chemosphere Glob Chang Sci 2000;2(34):397412. Joumard R, Jost P, Hickman J, Hassel D. Hot passenger car emissions modelling as a function of instantaneous speed and acceleration. Sci Total Environ 1995;169:16774. Joumard R, Andr M, Vidon R, Tassel P, Pruvost C. Influence of driving cycles on unit emissions from passenger cars. Atmos Environ 2000;34:46218. Kalligeros S, Zannikos F, Stournas S, Lois E, Anastopoulos G, Teas C, et al. An investigation of using biodiesel/marine diesel blends on the performance of a stationary diesel engine. Biomass Bioenergy 2003;24:1419. Kauman WG, Gottstein JW, Lantican D. Quality evaluation by numerical and subjective methods with application to dried veneer. Biometrics 1956;12:12753. Ketzel M, Whlin P, Berkowicz R, Palmgren F. Particle and trace gas emission factors under urban driving conditions in Copenhagen based on street and roof-level observations. Atmos Environ 2003;37 (20):273549. Kittelson DB, Watts WF, Johnson JP. Nanoparticle emissions on Minnesota highways. Atmos Environ 2004;38:919. Knothe G, Bagby MO, Ryan III TW, Callahan TJ, Wheeler HG. Vegetable oils as alternative diesel fuels: degradation of pure triglycerides during the precombustion phase in a reactor simulating a diesel engine. SAE Technical Paper Series. Warrendale, PA: Society of Automotive Engineers, Inc.; 1992. Paper no 920194. Kristensson A, Johansson C, Westerholm R, Swietlicki E, Gidhagen L, Wideqvist U, et al. Real-world traffic emission factors of gases and particles measured in a road tunnel in Stockholm, Sweden. Atmos Environ 2004;38(5):65773. Li H, Andrews GE, Zhu G, Daham B, Bell MC, Tate JE, et al. Impact of ambient temperature on exhaust thermal characteristics during cold start for real world SI car urban driving tests. SAE Technical Paper Series; 2005. paper no 2005-01-3896.
284
K. Ropkins et al. / Science of the Total Environment 376 (2007) 267284 emission levels. 2004 SAE World Congress, Detroit, Michigan, March 811; 2004. Schumacher LG, Borgelt SC, Fosseen D, Goetz W, Hires WG. Heavyduty engine exhaust emission tests using methyl ester soybean oil/ diesel fuel blends. Bioresour Technol 1996;57:316. Schurmann D, Staab J. On-the-road measurements of automotive emissions. Sci Total Environ 1990;93:14757. Sjdin A, Lenner M. On-road measurements of single vehicle pollutant emissions, speed and acceleration for large fleets of vehicles in different traffic environments. Sci Total Environ 1995;169(13):15765. Srivastava A, Prasad R. Triglycerides-based diesel fuels. Renew Sustain Energy Rev 2000;4:11133. Takada Y, Miyazaki T, Iida N. Study on local air pollution caused by NOx from diesel freight vehicle. SAE Technical Paper Series 200201-0651. Detroit, MI: Society of Automotive Engineers, Inc.; 2002. Turrio-Baldassarri L, Battistelli CL, Conti L, Crebelli R, De Berardis B, Iamiceli AL, et al. Emission comparison of urban bus engine fueled with diesel oil and biodiesel blend. Sci Total Environ 2004;327(13):14762. Watson HC. Effects of a wide range of drive cycles on the emissions from vehicles of three levels of technology. SAE Technical Paper Series; 1995. paper no 95-0221. Weilenmann M, Soltic P, Ajtay D. Describing and compensating gas transport dynamics for accurate instantaneous emission measurement. Atmos Environ 2003;37:513745.
Ma F, Hanna MA. Biodiesel production: a review. Bioresour Technol 1999;70:l15. Monyem A, Van Gerpen JH. The effect of biodiesel oxidation on engine performance and emissions. Biomass Bioenergy 2001;20: 31725. Nakamura H, Kihara N, Adachi M, Ishida K. Development of a wetbased NDIR and its application to On-Board Emissions Measurement System. SAE Technical Paper Series. Detroit, MI: Society of Automotive Engineers, Inc.; 2002. Paper no 2002-01-0612. Peterson CL, Reece DL, Thompson JC, Beck SM, Chases C. Ethyl ester of rapeseed used as a biodiesel fuel a case study. Biomass Bioenergy 1996;10(5/6):3316. Pierson WR, Gertler AW, Bradow RL. Comparison of the SCAQS tunnel study with other on-road vehicle emission data. J Air Waste Manage Assoc 1990;40:1495504. Pierson WR, Gertler AW, Robinson NF, Sagebiel JC, Zielinska B, Bishop GA, et al. Real-world automotive emissions Summary of studies in the Fort McHenry and Tuscarora mountain tunnels. Atmos Environ 1996;30(12):223356. Raheman H, Phadatare AG. Diesel engine emissions and performance from blends of karanja methyl ester and diesel. Biomass Bioenergy 2004;27:3937. Ramadhas AS, Jayaraj S, Muraleedharan C. Use of vegetable oils as I.C. engine fuels A review. Renew Energy 2004;29:72742. Ross M. Automobile fuel consumption and emissions: effect of vehicle and driving characteristics. Annu Rev Energy Environ 1994;19: 75112. Samuel S, Morrey D, Fowkes M, Taylor DHC, Austin L, Felstead T, et al. The most significant vehicle operating parameter for real-world