Effect of Extreme Temperatures and Driving Conditions on Gaseous Pollutants of a Euro 6d-Temp Gasoline Vehicle

: Gaseous emissions of modern Euro 6d vehicles, when tested within real driving emissions (RDE) boundaries, are, in most cases, at low levels. There are concerns, though, about their emission performance when tested at or above the boundaries of ambient and driving conditions requirements of RDE regulations. In this study, a Euro 6d-Temp gasoline direct injection (GDI) vehicle with three-way catalyst and gasoline particulate ﬁlter was tested on the road and in a laboratory at temperatures ranging between − 30 ◦ C and 50 ◦ C, with cycles simulating urban congested trafﬁc, uphill driving while towing a trailer at 85% of the vehicle’s maximum payload, and dynamic driving. The vehicle respected the Euro 6 emission limits, even though they were not applicable to the speciﬁc cycles, which were outside of the RDE environmental and trip boundary conditions. Most of the emissions were produced during cold starts and at low ambient temperatures. Heavy trafﬁc, dynamic driving, and high payload were found to increase emissions depending on the pollutant. Even though this car was one of the lowest emitting cars found in the literature, the proposed future Euro 7 limits will require a further decrease in cold start emissions in order to ensure low emission levels under most ambient and driving conditions, particularly in urban environments. Nevertheless, motorway emissions will also have to be controlled well.


Introduction
Air pollution, specifically particulate matter (PM), nitrogen dioxide (NO 2 ), and ground-level ozone (O 3 ), has significant impacts on the health of the European population, particularly in urban areas [1]. The lockdown measures introduced by most European countries to reduce the transmission of COVID-19 in the spring of 2020 led to significant reductions in the emission of air pollutants, particularly from road transport, aviation, and international shipping. In particular, NO 2 concentrations were significantly reduced, independent of meteorological conditions [2][3][4][5][6], highlighting the important contribution of road transport to air quality, particularly in cities.
The winter season and low temperatures are usually correlated with high ambient air pollution due to increased emissions (household heating, coal burning, and road transport being the main contributors) [7][8][9][10]. Weather conditions that limit the advection and • WLTC (worldwide harmonized light vehicles test cycle) Type 1 approval cycle. As urban part, the low and medium phases were considered, according to the RDE regulation, while the extra high phase was used as motorway part [44]. • TfL (Transport for London urban interpeak) urban driving characterized by stop and go traffic in congested conditions. The cycle was developed by Millbrook Inc. in collaboration with the Traffic for London Authority [45]. • BAB 130 (Bundesautobahn, Federal highway) high speed driving on the motorway up to 130 km/h, with frequent and sharp accelerations. It was developed by ADAC (Allgemeiner Deutscher Automobil-Club e.V.) as part of the EcoTest car testing protocol [46]. • RDE short: A one-hour duration test with urban (time share 53%), rural (28%), and motorway conditions (19%), and road slope (range −9.6% to 9.2%) was provided by Ricardo Automotive & Industrial. • RDE boundary: A two-hour cycle recreating the most dynamic drive possible within the RDE boundaries with a 90% payload, including road slope (range −8.1% to 6.5%), provided by FEV Europe. The urban/rural/motorway time shares were 66%/20%/15%. • Uphill: A cycle (vehicle speed < 60 km/h) simulating (i) uphill driving with a 5% constant slope, while towing a 800 kg trailer (uphill tow); (ii) uphill driving with a 5% constant slope, car loaded to 85% payload and towing a 1700 kg trailer (85% of max trailer weight) (uphill tow 85%). The cycle was based on actual uphill driving data at the JRC premises. • RDE road: Two different routes according to Type 1A on-road procedure (RDE road) with a portable emissions measurement system (PEMS) (MOVE from AVL); routes were actually driven on the road at the JRC premises. The WLTC tests were conducted using the test mass and roadload coefficients declared on the CoC (Certificate of Conformity) of the vehicle, after roadload derivation on the dyno. For the other cycles, the dyno coefficients were adjusted depending on the simulated conditions (e.g., slope, extra weight, etc.). The 95th percentile of the product of vehicle speed per positive acceleration greater than 0.1 m/s 2 (v × a) for urban, rural, and motorway shares is plotted in Figure 1. • RDE road: Two different routes according to Type 1A on-road procedure (RDE road) with a portable emissions measurement system (PEMS) (MOVE from AVL); routes were actually driven on the road at the JRC premises.  (%) No No 5% 5% −9.6 to 9.2% −8.1 to 6.5% −7.3 to 9.2% −9.8 to 10.6% No The WLTC tests were conducted using the test mass and roadload coefficients declared on the CoC (Certificate of Conformity) of the vehicle, after roadload derivation on the dyno. For the other cycles, the dyno coefficients were adjusted depending on the simulated conditions (e.g., slope, extra weight, etc.). The 95th percentile of the product of vehicle speed per positive acceleration greater than 0.1 m/s 2 (v × a) for urban, rural, and motorway shares is plotted in Figure 1. Ambient temperatures tested were −30 °C, −10 °C, −7 °C, 5 °C, 23 °C, and 50 °C for the TfL, BAB, and RDE short cycles. The RDE boundary and the uphill cycles were driven at −7 °C and/or −10 °C, as these were considered the most challenging situations at or slightly above the permitted RDE boundary conditions. The WLTC test temperature was 23 °C, as required in the regulation, while the ambient temperature during the on-road tests was 1720 °C. Temperatures of −30 °C and 50 °C were selected because these conditions are far from the normal operative conditions for which the vehicle might be calibrated. Note that such temperatures are not expected in Europe and compliance with any limits is highly unlikely. Ambient temperatures tested were −30 • C, −10 • C, −7 • C, 5 • C, 23 • C, and 50 • C for the TfL, BAB, and RDE short cycles. The RDE boundary and the uphill cycles were driven at −7 • C and/or −10 • C, as these were considered the most challenging situations at or slightly above the permitted RDE boundary conditions. The WLTC test temperature was 23 • C, as required in the regulation, while the ambient temperature during the on-road tests was 1720 • C. Temperatures of −30 • C and 50 • C were selected because these conditions are far from the normal operative conditions for which the vehicle might be calibrated. Note that such temperatures are not expected in Europe and compliance with any limits is highly unlikely.

Results
Initially, the laboratory type-approval Type 1 results are given, followed by realtime examples. Special attention will be given to cold start emissions. Then, urban and motorway emissions will be presented. Our focus is on regulated pollutants (CO, NO x , Atmosphere 2021, 12, 1011 5 of 25 THC) and only NH 3 from non-regulated pollutants. Other non-regulated pollutants were low for the specific vehicle (e.g., N 2 O < 3 mg/km). The emissions per km or kg of fuel can be found in the Supplementary Material.
3.1. Type 1 (WLTC) and Type 1A (RDE) Emissions Table 2 summarizes the results for the on-road Type 1A (RDE), the laboratory Type 1 (WLTC), and the laboratory RDE-like cycles at different ambient temperatures. The WLTC emissions were three to six times lower than the respective Euro 6 limits applicable to the specific Euro 6d-Temp vehicle. The official tests (RDE road) were in agreement with the laboratory tests and much below the respective limit for NO x , where a conformity factor of 2.1 would be applicable for the specific vehicle to take into account the PEMS measurement uncertainty [47]. Various short (1 h) RDE-like cycles, which were not RDE compliant due to their short duration, at different temperatures also had emissions below the respective limits (although not applicable). The only exception was the −30 • C test, which had CO just at the Euro 6 limit, and THC 27% higher than the applicable limit at 23 • C. All these tests at −30 • C and −10 • C (and 50 • C) were outside the boundaries of the RDE regulation. The RDE at the boundaries (−10 • C, 90% payload, and dynamic driving at the limits of RDE) was also within the emission limits.
It should be mentioned that a correction of 1.6 would be applicable to emissions on RDE cycles at extended conditions (−7 • C to 0 • C, 30 • C to 35 • C) if the emissions calculation method was conducted according to RDE regulation (EU) 2017/1151. This correction was not applied to our analysis. Furthermore, an "RDE evaluation factor" should be applied at each test. It is calculated as the ratio of the test (trip) CO 2 and the type-approval cycle CO 2 , and takes into account the severity of the trip. For ratios < 1.3, no correction is applied while, for ratios > 1.5, the inverse of the ratio is applied to the final emission result. For the specific vehicle, which was type-approved before January 2020, the respective ratios were 1.2 and 1.25. Again, no such corrections were applied to the results presented in this study. With such corrections, the results of the −30 • C would be well within the emission limits.

Real Time Examples
As an example, Figure 2 provides the speed profile and the cumulative emissions for the TfL (Transport for London) and BAB (Bundesautobahn, Federal highway) cycles for various ambient temperatures. More information regarding coolant temperatures can be found in Appendix A Figure A1. The temperature profiles indicate a different engine thermal management strategy after 1700 s for the 50 • C test. There was no additional infor- mation available about air-to-fuel ratios or TWC temperatures. Exhaust gas temperatures can be found in the companion paper.

Real Time Examples
As an example, Figure 2 provides the speed profile and the cumulative emissions for the TfL (Transport for London) and BAB (Bundesautobahn, Federal highway) cycles for various ambient temperatures. More information regarding coolant temperatures can be found in Appendix Figure A1. The temperature profiles indicate a different engine thermal management strategy after 1700 s for the 50 °C test. There was no additional information available about air-to-fuel ratios or TWC temperatures. Exhaust gas temperatures can be found in the companion paper.    temperatures (−10 and −7 • C), more than the cold start emissions of mild temperatures (23 and 5 • C), and substantially more than the cold start emissions at high temperature (50 • C). Similar behavior was also seen for THC (Figure 2b) and NO x (Figure 2c). While ambient temperature had a clear effect on cold start emissions for CO and THC, this trend was not clear for NO x . For CO and NO x , there was a clear increase in emissions during accelerations, but not for THC. NH 3 emissions ( Figure 2d) followed a similar trend. Although there was an increase during cold start, the step increase was noticed at 200 s, after the catalyst had reached its light-off temperature (as indicated by the stabilization of the cumulative emissions for the regulated pollutants). Increased emissions during the accelerations of the BAB cycle were also noticed.
In general, cumulative emissions stabilized much faster than the coolant temperature (compare with Figure A1), indicating that the TWC reached its efficient operating temperature faster than the engine. Of particular interest was the increase in emissions (in particular CO and NH 3 ) at the last part of the 50 • C TfL test (around 1700 s), which coincided with the coolant temperature decrease which indicated a different engine thermal management.
The trends were similar for the RDE short cycles. After the cold start increase, cumulative emissions were flat (or smoothly increasing) at the urban part, and then increased slightly at the rural part and increased significantly at the motorway part. Examples are given in Appendix A Figures A2 and A3. Figure 3 plots the cumulative emissions during the first 15 min for the various urban cycles. Figure A4 in the Appendix A plots the coolant temperatures (no TWC temperatures were available). Most of the emissions of the urban part took place during the first one-two minutes, with a few exceptions. For the TfL at −7 • C, it took three minutes to reach a "plateau" in CO cumulative emissions (Figure 3a). Another notable exception was uphill driving with a simulated trailer with an 85% of maximum payload, where the CO emissions continued to increase almost linearly after the first two minutes (Figure 3a). The dynamic RDE boundary had the highest cold start CO emissions. THC emissions (Figure 3b) showed a similar trend, with the exception that uphill driving did not further increase the emissions after the first minute. Similarly, for NO x , after the first minute, the cumulative emissions were constant (Figure 3c). Uphill and dynamic driving had similarly high emissions, while the short RDE was the lowest. For NH 3 , as mentioned before, the increase appeared after the catalyst light-off, with the highest increase taking place somewhere between 80 s and 180 s (Figure 3d). The emissions continued to increase after the first three minutes. The highest emissions were noticed in uphill high load driving, followed by the dynamic cycle. Figure 2a shows cumulative CO emissions. CO cold start emissions were higher for lower ambient temperatures. After the first minute (three minutes for −30 °C), the cumulative emissions remained almost constant until the motorway BAB cycles. The dynamic accelerations of the BAB contributed substantially less than the cold start emissions of the extreme low temperature (−30 °C), were comparable to the cold start emissions of low temperatures (−10 and −7 °C), more than the cold start emissions of mild temperatures (23 and 5 °C), and substantially more than the cold start emissions at high temperature (50 °C). Similar behavior was also seen for THC (Figure 2b) and NOx (Figure 2c). While ambient temperature had a clear effect on cold start emissions for CO and THC, this trend was not clear for NOx. For CO and NOx, there was a clear increase in emissions during accelerations, but not for THC.
NH3 emissions ( Figure 2d) followed a similar trend. Although there was an increase during cold start, the step increase was noticed at 200 s, after the catalyst had reached its light-off temperature (as indicated by the stabilization of the cumulative emissions for the regulated pollutants). Increased emissions during the accelerations of the BAB cycle were also noticed.
In general, cumulative emissions stabilized much faster than the coolant temperature (compare with Figure A1), indicating that the TWC reached its efficient operating temperature faster than the engine. Of particular interest was the increase in emissions (in particular CO and NH3) at the last part of the 50 °C TfL test (around 1700 s), which coincided with the coolant temperature decrease which indicated a different engine thermal management.
The trends were similar for the RDE short cycles. After the cold start increase, cumulative emissions were flat (or smoothly increasing) at the urban part, and then increased slightly at the rural part and increased significantly at the motorway part. Examples are given in Appendix Figures A2 and A3. Figure 3 plots the cumulative emissions during the first 15 min for the various urban cycles. Figure A4 in the Appendix plots the coolant temperatures (no TWC temperatures were available). Most of the emissions of the urban part took place during the first onetwo minutes, with a few exceptions. For the TfL at −7 °C, it took three minutes to reach a "plateau" in CO cumulative emissions (Figure 3a). Another notable exception was uphill driving with a simulated trailer with an 85% of maximum payload, where the CO emissions continued to increase almost linearly after the first two minutes (Figure 3a). The dynamic RDE boundary had the highest cold start CO emissions. THC emissions (Figure 3b) showed a similar trend, with the exception that uphill driving did not further increase the emissions after the first minute. Similarly, for NOx, after the first minute, the cumulative emissions were constant (Figure 3c). Uphill and dynamic driving had similarly high emissions, while the short RDE was the lowest. For NH3, as mentioned before, the increase appeared after the catalyst light-off, with the highest increase taking place somewhere between 80 s and 180 s (Figure 3d). The emissions continued to increase after the first three minutes. The highest emissions were noticed in uphill high load driving, followed by the dynamic cycle.
(a)   Figure 4 summarizes the cumulative emissions (in g) during the first 300 s of cold start test cycles at various ambient temperatures. A duration of 300 s was selected in order to ensure the "plateau" region was reached (i.e., a small increase in the emissions afterwards), even in the extreme case of the −30°C for the great majority of test cycles and pollutants, as presented in the previous section. For the other temperatures, the plateau region was reached within one to two minutes (see also Figures 2 and 3). Furthermore, a cold start in the RDE regulation is defined as the first 300 s. The distance driven in the first 300 s varied from 1 km (TfL) to 2.8 km (uphill and RDE boundary). In the case of the RDE, the distance covered was 1.7 km.

Ambient Temperature and Cold Start Emissions
There is a clear increasing trend with decreasing ambient temperature for CO and THC emissions (Figure 4a,b). The cold start (i.e., first 5 min) CO mass was <4.5 g at temperatures < 5 °C, 10-18 g at −10 °C and 32-40 g at −30 °C. For the same temperature (−10 °C), the dynamic RDE boundary cycle had the highest emissions.
The cold start THC mass was <1 g at temperatures above 5 °C, 2-4 g at −10 °C and 7 g at −30 °C. For the same temperature (−10 °C), the dynamic RDE boundary cycle had the highest emissions. At low ambient temperatures, the TfL had higher emissions than the RDE short.  Figure 4 summarizes the cumulative emissions (in g) during the first 300 s of cold start test cycles at various ambient temperatures. A duration of 300 s was selected in order to ensure the "plateau" region was reached (i.e., a small increase in the emissions afterwards), even in the extreme case of the −30 • C for the great majority of test cycles and pollutants, as presented in the previous section. For the other temperatures, the plateau region was reached within one to two minutes (see also Figures 2 and 3). Furthermore, a cold start in the RDE regulation is defined as the first 300 s. The distance driven in the first 300 s varied from 1 km (TfL) to 2.8 km (uphill and RDE boundary). In the case of the RDE, the distance covered was 1.7 km.

Ambient Temperature and Cold Start Emissions
There is a clear increasing trend with decreasing ambient temperature for CO and THC emissions (Figure 4a,b). The cold start (i.e., first 5 min) CO mass was <4.5 g at temperatures <5 • C, 10-18 g at −10 • C and 32-40 g at −30 • C. For the same temperature (−10 • C), the dynamic RDE boundary cycle had the highest emissions.
The cold start THC mass was <1 g at temperatures above 5 • C, 2-4 g at −10 • C and 7 g at −30 • C. For the same temperature (−10 • C), the dynamic RDE boundary cycle had the highest emissions. At low ambient temperatures, the TfL had higher emissions than the RDE short.  For NOx, the trend in function of temperature was less evident, and the emissions had a bathtub-like curve, especially in the case of the TfL cycle; high emissions were seen at both low and high temperatures, and the TfL was higher than the RDE short (Figure 4c). The dynamic RDE boundary and the high load uphill cycles also showed high emissions. Comparing the CO and NOx graphs (e.g., for TfL), the NOx-CO trade-off related to lambda control could be seen; when CO was high, NOx was usually low, and vice versa.
For NH3 (Figure 4d), the temperature dependence was almost non-existent. For a specific temperature, there was an influence of the cycle: (i) the dynamic cycle had high emissions; (ii) the stop and go TfL had higher emissions than the moderate RDE short; (iii) the high load uphill cycle had high emissions (~140 mg/km over the first 2.8 km). Figure 5 presents the emissions (in mg/km) of the vehicle for the urban cycles and urban parts of the rest cycles. The TfL was an 8.9 km urban cycle and the urban part of the RDE short urban cycle was 12.7 km, while the urban parts of the dynamic RDE boundary and actual RDE on-road cycles were around 35 km long. The uphill cycles were 9 km long, representing uphill driving towing a trailer with 85% of the maximum weight. The WLTC low and medium part were 7.8 km long. It should be recalled that the minimum urban distance in the current regulations is 16 km; thus, any comparison with limits of non-RDE compliant tests (e.g., TfL, uphill) is only for illustrative reasons. For NO x , the trend in function of temperature was less evident, and the emissions had a bathtub-like curve, especially in the case of the TfL cycle; high emissions were seen at both low and high temperatures, and the TfL was higher than the RDE short (Figure 4c). The dynamic RDE boundary and the high load uphill cycles also showed high emissions. Comparing the CO and NO x graphs (e.g., for TfL), the NO x -CO trade-off related to lambda control could be seen; when CO was high, NO x was usually low, and vice versa.

Ambient Temperature and Urban Emissions
For NH 3 (Figure 4d), the temperature dependence was almost non-existent. For a specific temperature, there was an influence of the cycle: (i) the dynamic cycle had high emissions; (ii) the stop and go TfL had higher emissions than the moderate RDE short; (iii) the high load uphill cycle had high emissions (~140 mg/km over the first 2.8 km). Figure 5 presents the emissions (in mg/km) of the vehicle for the urban cycles and urban parts of the rest cycles. The TfL was an 8.9 km urban cycle and the urban part of the RDE short urban cycle was 12.7 km, while the urban parts of the dynamic RDE boundary and actual RDE on-road cycles were around 35 km long. The uphill cycles were 9 km long, representing uphill driving towing a trailer with 85% of the maximum weight. The WLTC low and medium part were 7.8 km long. It should be recalled that the minimum urban distance in the current regulations is 16 km; thus, any comparison with limits of non-RDE compliant tests (e.g., TfL, uphill) is only for illustrative reasons. Atmosphere 2021, 12, x FOR PEER REVIEW 10 of 25  To put the results into context, the RDE limits were also plotted for moderate conditions (0 °C to 30 °C) (shown as a green box). At the extended conditions of the RDE (−7 °C to 0 °C, 30 °C to 35 °C), a 1.6 correction factor should be applied to the emissions. Here, instead of reducing the emissions with this factor, we increased the moderate limit (adjusted limits shown as red boxes) for better visualization and to leave the actual emissions of the vehicle intact. In addition to this correction to the emission limits (red box), the RDE evaluation factors (CO2 corrections) were plotted as purple boxes. For urban trips the onroad, CO2 was divided by the CO2 of the L + M phases of the WLTC. For our tests, the ratios were 1.4-1.9 for the RDE short urban and 1.7-2.5 for the TfL. The higher values were for the low temperature and high temperature tests. For the uphill driving cycles, the ratios were 2.8-3.5, confirming the high-power demand of the specific cycles.

Ambient Temperature and Urban Emissions
For NOx measured on the road with PEMS, a conformity factor of 2.1 was applicable to the specific Euro 6d-Temp vehicle [47]. The conformity factor for Euro 6d vehicles is 1.43 and will be lowered in the future. The other gaseous pollutants have no conformity factors because they are not yet regulated and verified on the road. As all our tests were conducted on the chassis dynamometer with laboratory grade equipment (except RDE  Table 2. Corrections are also given in purple boxes. No conformity factors (for the PEMS uncertainty) were applied to the limits. The grey triangle for the RDE boundary cycle refers to the first 10 km, while the non-shaded triangle corresponds to the complete urban part (36 km) of the cycle: (a) CO; (b) THC; (c) NO x ; (d) NH 3 . Dotted boxes with limits on CO and THC emissions are plotted for illustration purposes only, as no on-road limit for those species is applicable for Euro 6 vehicles.
To put the results into context, the RDE limits were also plotted for moderate conditions (0 • C to 30 • C) (shown as a green box). At the extended conditions of the RDE (−7 • C to 0 • C, 30 • C to 35 • C), a 1.6 correction factor should be applied to the emissions. Here, instead of reducing the emissions with this factor, we increased the moderate limit (adjusted limits shown as red boxes) for better visualization and to leave the actual emissions of the vehicle intact. In addition to this correction to the emission limits (red box), the RDE evaluation factors (CO 2 corrections) were plotted as purple boxes. For urban trips the on-road, CO 2 was divided by the CO 2 of the L + M phases of the WLTC. For our tests, the ratios were 1.4-1.9 for the RDE short urban and 1.7-2.5 for the TfL. The higher values were for the low temperature and high temperature tests. For the uphill driving cycles, the ratios were 2.8-3.5, confirming the high-power demand of the specific cycles.
For NO x measured on the road with PEMS, a conformity factor of 2.1 was applicable to the specific Euro 6d-Temp vehicle [47]. The conformity factor for Euro 6d vehicles is 1.43 and will be lowered in the future. The other gaseous pollutants have no conformity factors because they are not yet regulated and verified on the road. As all our tests were conducted on the chassis dynamometer with laboratory grade equipment (except RDE road), no conformity factor was applied, and all results were compared with the laboratory limits.
The results are quite similar to Figure 4. The reason for this is that, for the urban section of the tests, most of the emissions are produced during the first minutes, so any difference is mainly due to the different length of the urban cycle.
The CO emissions (Figure 5a) from 550 mg/km at moderate conditions increased up to 2400 mg/km at the uphill cycles at −7 • C and 4700 mg/km at the TfL at −30 • C. At 50 • C, CO was 1100 mg/km. The TfL emissions were, in general, higher than the short RDE, with the differences increasing at the extreme temperatures. The RDE boundary was lower than the rest cycles, but the urban distance was three times longer. Considering only the first 10 km, the emissions were 2000 mg/km, close to the uphill cycles, which had the highest emissions at the same ambient temperature (−7 • C or −10 • C).
The THC emissions (Figure 5b) showed an increasing trend with decreasing ambient temperature. From 50 mg/km at 50 • C and 50-150 mg/km at 5-23 • C, they reached 150-400 mg/km at −10 • C and 500-800 mg/km at −30 • C. The loaded uphill cycles were between the TfL and short RDE, while the RDE boundary had the lowest emissions for the same temperature (or the highest when considering only the first 10 km). CH 4 emissions (no figure shown) were generally <10% of THC emissions (20% for THC <20 mg/km). Percentages around 30% were measured for the RDE boundary and TfL at positive ambient temperatures.
The NO x emissions showed a bathtub-like curve in function of the ambient temperature (Figure 5c). They ranged from 10 mg/km to 35 mg/km at the 5-23 • C region, and up to 70-110 mg/km at −10 • C or 50 • C on the TfL cycle. The high load uphill cycles at −10 • C were also around 80 mg/km. However, emissions were kept at 30 mg/km at −10 • C for the urban part of the short RDE and the dynamic RDE (but at 75 mg/km considering only the first 10 km). Interestingly, they were <55 mg/km at −30 • C for the same cycle (RDE short).
The NH 3 emissions (Figure 5d) were relatively constant for different temperatures. High ammonia emissions (30 mg/km) were measured at the TfL, irrespective of ambient temperature and the uphill towing cycles (40-60 mg/km). The RDE boundary also had 30 mg/km when only the first 10 km were considered. Lower levels (10-20 mg/km) were measured with the RDE short. Currently, there are no limits on this pollutant. For TfL, the average 35 mg/km NH 3 emissions resulted from slightly less than 20 ppm average NH 3 concentrations, with peaks of 200-400 ppm during the first minutes.
The N 2 O emissions (no figure shown) were <3 mg/km, except for the TfL tests, which were around 5 mg/km; these were always confined to the temperature window of the catalyst light-off. CO emissions (Figure 6a) were approximately 400 mg/km, with the exception of the RDE boundary at −10 • C and the cycles at 50 • C, where the emissions reached 1000 mg/km (still below the laboratory limit, even though they were not applicable). The THC (Figure 6b) were below 20 mg/km in all tests, except at the RDE boundary, where they reached 35 mg/km. At the RDE boundary, THC spikes were evident in the motorway part, but the emissions were still three times below the laboratory limit of 100 mg/km. The NO x emissions (Figure 6c) were only 5 mg/km, reaching 20 mg/km in the BAB cycle at 50 • C. The dynamic RDE boundary at −10 • C had 15 mg/km. All results were lower than the (non-applicable) limit of 60 mg/km. The NH 3 emissions (Figure 6d

Discussion
This study assessed the emissions of a Euro 6d-Temp gasoline vehicle with a closecoupled TWC and an underfloor uncoated GPF, focusing on extreme temperatures and driving conditions. The results confirmed previous findings, i.e., (i) during urban operation, the majority of emissions comes from cold starts; (ii) the lower the ambient temperature, the higher the cold start emissions, namely for CO and THC (not so evident for NH3); (iii) dynamic driving and/or high engine load increase the emissions of most pollutants. The major contribution of this study is the extension of these findings to extreme ambient temperatures (−30 °C and 50 °C) and driving conditions (traffic, dynamic driving, trailer towing).

Cold Start
Higher emissions during a cold start are known and attributed to higher engine-out emissions and low efficiency of the aftertreatment devices [48][49][50]. It was shown that, even though it took long time for the coolant temperature to stabilize, the cumulative emissions, in most cases, stabilized within a few minutes; this highlights the importance of aftertreatment devices' status on tailpipe emissions, which is in agreement with other studies [51]. Cold start emissions, defined here as emissions in the first 300 s (and not the difference

Discussion
This study assessed the emissions of a Euro 6d-Temp gasoline vehicle with a closecoupled TWC and an underfloor uncoated GPF, focusing on extreme temperatures and driving conditions. The results confirmed previous findings, i.e., (i) during urban operation, the majority of emissions comes from cold starts; (ii) the lower the ambient temperature, the higher the cold start emissions, namely for CO and THC (not so evident for NH 3 ); (iii) dynamic driving and/or high engine load increase the emissions of most pollutants. The major contribution of this study is the extension of these findings to extreme ambient temperatures (−30 • C and 50 • C) and driving conditions (traffic, dynamic driving, trailer towing).

Cold Start
Higher emissions during a cold start are known and attributed to higher engine-out emissions and low efficiency of the aftertreatment devices [48][49][50]. It was shown that, even though it took long time for the coolant temperature to stabilize, the cumulative emissions, in most cases, stabilized within a few minutes; this highlights the importance of aftertreatment devices' status on tailpipe emissions, which is in agreement with other studies [51]. Cold start emissions, defined here as emissions in the first 300 s (and not the difference between cold and hot start emissions [52]), contributed, in many cases, >80% of CO, THC, and NO x urban emissions (Table 3). For CO, the contribution was lower in the 50 • C tests and the high load uphill towing cycles because there were also considerable emissions during the rest of the urban phase. For NO x , the contribution was lower in most RDE short tests. Other studies have also found a high contribution from cold start for CO (>70%), but this is smaller for NO x (40-60%) for turbo-charged GDI vehicles [53][54][55]; however, lower percentages have also been reported for RDE trips [32,56]. Table 3. Contribution of cold start (300 s) to total urban emissions, calculated as g over total g. Temp. is the ambient temperature. Distance is the cold start and urban distance. Time is the urban trip duration.

Cycle (Urban Part)
Temp The cold start contribution was around 50% for NH 3 . NH 3 has been shown to primarily form within the exhaust temperature range of 250-550 • C [57]. NH 3 formation over the catalyst is enhanced at low air/fuel (lambda) ratios (rich operating conditions). The mechanisms that take place under these conditions are the water-gas (CO) shift reaction, producing H 2 and the reaction of NO and H 2 [58]. The availability of H 2 and CO, the exhaust gas temperature, and the lambda also affects small differences in the slopes of the cumulative NH 3 functions between different temperatures [59]. Table 4 summarizes the contribution of a cold start to total RDE trips. In general, the contribution was <50% for CO and NO x , <25% for NH 3 , but >50% for THC. Higher percentages were measured at lower ambient temperatures. For the actual on-road tests, the contribution was 6-19% for CO and NO x , which is in agreement with other studies [56,60].

Urban Emissions
The urban emissions of the regulated pollutants (CO, THC, NO x ) were primarily determined by cold start emissions. As the contribution of the hot urban operation was small for most pollutants, a longer distance resulted in lower distance specific emissions. The emission factors presented in Figure 5 were quite comparable because the urban distances were within 9-12 km. Figure 7a presents the emissions in function of distance.
CO is plotted as an example. Each cycle is plotted with a different color. Different groups of curves can be seen; the highest emissions are at −30 • C, followed by the −7 • C and −10 • C tests. The other group is the 5 • C, 23 • C, and 50 • C tests. For each group of temperatures, TfL emissions are usually higher than the RDE short (primarily NO x and NH 3 ). RDE boundary emissions are higher than the other cycles; the RDE road and WLTC are lower than the rest cycles. What is important from this figure is the decreasing trend in emissions (note the log scale) in function of distance, with a few exceptions: the uphill cycles and the 50 • C tests, as discussed before (Figure 3). Thus, for pollutants for which the cold start contribution is important, total distance of the urban cycle is also important. For NO x (Figure 8a), the highest emissions were noted for uphill, RDE boundary, and TfL (all cycles at −7 • C or −10 • C), as well as TfL at 50 • C. For NH 3 , the highest emissions were seen for uphill with an 85% payload (car and trailer) (Figure 8b).  Many studies plot emissions in function of the average speed [29,61] because, typically, a decreasing trend is expected for low average speeds. As an example, Figure 7b plots CO emissions in function of the average speed for different urban cycles. The typical curve, where emissions increase as average speed decreases, is not so evident. Ambient temperature, dynamic driving, high load, and total distance all have a bigger influence. For the other pollutants, the results are similar, and the average speed dependence is even smaller. Such a lack of correlation was also shown recently for vehicles fulfilling China 3 to China 5 emission limits [62].

Ambient Temperature
Higher emissions due to lower ambient temperatures are well documented [22,30,[63][64][65]. High viscosity and friction are the primary causes of emission increases; excess fueling is also important [66]. The use of air conditioning can also contribute to an increase in emissions, even at low temperatures, due to the increase in engine load (and possibly richer fuel injection, injection pressure, and delayed spark timing) needed to operate the compressor and/or fan of the air conditioning [67,68]. In our study, the air conditioning of the vehicle was always enabled at 21.5 °C. At lower ambient temperatures, the TWC can take longer to reach the light-off temperature [29] and/or the lambda is not kept in unity. In this study, it was shown that, even in the extreme temperature of −30 °C and the challenging traffic cycle of TfL, the TWC could operate efficiently within <300 s. Table 5 summarizes the emission ratios at various temperatures compared to 23 °C for TfL (cold start, urban), BAB (motorway), and the RDE short (cold start, urban, motorway).

Ambient Temperature
Higher emissions due to lower ambient temperatures are well documented [22,30,[63][64][65]. High viscosity and friction are the primary causes of emission increases; excess fueling is also important [66]. The use of air conditioning can also contribute to an increase in emissions, even at low temperatures, due to the increase in engine load (and possibly richer fuel injection, injection pressure, and delayed spark timing) needed to operate the compressor and/or fan of the air conditioning [67,68]. In our study, the air conditioning of the vehicle was always enabled at 21.5 • C. At lower ambient temperatures, the TWC can take longer to reach the light-off temperature [29] and/or the lambda is not kept in unity. In this study, it was shown that, even in the extreme temperature of −30 • C and the challenging traffic cycle of TfL, the TWC could operate efficiently within <300 s. Table 5 summarizes the emission ratios at various temperatures compared to 23 • C for TfL (cold start, urban), BAB (motorway), and the RDE short (cold start, urban, motorway).
CO cold start emissions at −7 • C of Euro 4 vehicles have been shown to be 15-45 g [69]; cold start values are between 6 and 32 g for Euro 5 [34,70] and at an average of 27 g for China 6 GDI [10]. However, in our study, they were 10-15 g. At −20 • C, the cold start emissions of the Euro 4 vehicles were 50-160 g [69] while, in our study, they were 32-40 g at −30 • C. Distance specific emission factors of four Euro 6 GDI vehicles were 1.3-8.0 g/km (over the first phase of WLTC) [30], while four Euro 6b GDI vehicles emitted 5-12 g/km at the urban Artemis cycle at−7 • C [35]. These values are, in general, higher than the 1.5 g/km found in our study, confirming that the vehicle of our study was one of the best vehicles in terms of emissions. Both the absolute levels and the ratio of the −7 • C to 23 • C cold start emissions were on the lower side of reported values. In our study, this value was close to 3 (Table 5), while typical factors of 4-5 were reported [35,66,70], and sometimes even >8 [35,66,69,71]. Similarly, THC cold start emissions at−7 • C for Euro 4 vehicles were 5-15 g [69]; cold start values are between 1-4 g for Euro 5 [34,70] and at an average of 3 g for China 6 GDI [10]. However, in our study, they were 2-3 g. The emissions increased by a factor of 2.5-3 compared to 23 • C, which is again lower than what is reported in the literature [35,66].
On the other hand, no clear effect of ambient temperature on NO x for gasoline cars is reported in the literature [17,35,56,69,72]. However, some studies report a relatively small increase [10,30,34]: At −7 • C, NO x emissions were, on average, 1.7 times higher than at 23 • C [30]. Higher NO x emissions from spark ignition vehicles may be related to lower catalytic efficiency and longer periods to reach light-off temperature at cold ambient temperatures [30]. On the other hand, at lower temperatures, combustion is usually rich, which results in relatively low NO x emissions. With rich mixtures, the catalyst promotes a reduction in NO x by reactions involving HC and CO [73]. Thus, it seems emissions reduction is highly dependent on the strategy of a given vehicle's manufacturer. NH 3 emissions only marginally increased at lower ambient temperatures. This is in agreement with other studies, where NH 3 emissions from Euro 6 vehicles increased on average 1.5 times while temperature decreased from 23 to −7 • C [30]; however, much higher increases have also been reported [70], which are probably due to enrichment. At higher than 23 • C, a study found an increase in NH 3 emissions [74], while, in our tests, the increase was negligible. It seems that the parameters that increase NH 3 (temperature and lambda) were well controlled by the vehicle of our study.
Studies at high ambient temperatures are scarce [17,74,75]. Tests with Euro 3 vehicles with air-conditioning switched on showed a clear increase in CO and THC emissions at 37 • C, but not for NO x . Another study did not find any effect on CO and NO x at temperatures up to 40 • C [74]. Here, we found no effect on CO and HC cold start emissions, but a clear effect on NO x was apparent. That said, we documented an effect on CO emissions of the motorway cycles. The differences compared to older studies can be attributed to better lambda control at high speeds with the vehicle of our study (for CO and THC) and/or correctly sized TWC (adequate space velocities for the highest exhaust flows). More fundamental studies in that direction are necessary to generate a deeper understanding.

Dynamic Driving
Dynamic driving had a clear effect on CO and THC. Aggressive driving resulted in deviations from stoichiometric mixture, affecting both engine-out emissions and TWC conversion efficiency. The dynamic driving effect was evident at both cold start and motorway cycles. In the urban cycles, it was less evident because the RDE boundary distance was more than three times longer than the rest cycles, resulting in distance-specific emissions at the same levels of the rest cycles. However, when the dynamic RDE boundary was calculated for the first 10 km, the emissions were clearly high. Higher emissions with dynamic driving for CO have also been reported by others [23]; typically, emissions increased by a factor of 3, but some vehicles increased by a factor of 30 on urban routes.
For NO x and NH 3 , the effect of dynamic driving was smaller, which is in agreement with other studies [76]. For these two pollutants, emission levels of the dynamic cycle were similar to the TfL or uphill driving cycles. Similar, or even reduced, NO x emissions with dynamic driving have been reported by others [77]; however, these typically increase by around 80% on urban routes [23]. A study showed that, during fuel-cut events, air oxidizes the catalyst. Then, the excess fuel during accelerations results in NH 3 formation on the first part of the catalyst. When the rear part of the catalyst is not yet reduced, it oxidizes NH 3 back to NO x [78]. Another study showed that the instantaneous peaks of v × a had a clear one-to-one correspondence with the peaks of instantaneous NO x emissions [53]. For the same route, higher v × a results in higher emissions [23,79] but, when different routes are compared, this is not necessarily true [41] because v × a does not take into account slope or additional weight.

Towing and Uphill Driving
The cycle including uphill driving and towing a trailer with 85% of the maximum payload had high cold start emissions, particularly NH 3 . The reason for is high engine-out emissions (not for NH 3 ) and low conversion efficiency of the catalyst during the first minutes. In particular, for NH 3 , in the TWC, in addition to the NH 3 coming from a rich operation, the high exhaust gas temperature favored the formation of NH 3 from engineout NO. The increasing trend in cumulative emissions also continued after the cold start, indicating rich engine operation due to the high power demand, as the TWC had already reached its operating temperature. We are not aware of similar tests, especially at low ambient temperatures of −10 • C.

Motorway Emissions
Total CO and NH 3 presented non-negligible emissions during motorway operation. However, the motorway distance was higher than in urban operations, resulting in lower distance-specific emissions. Approximately 10 mg/km of NH 3 emissions corresponded to slightly less than 10 ppm of average NH 3 concentrations, with peaks between 50 and 100 ppm during accelerations. Similar levels have been reported by others [80]. In the case of NO x and THC, the contribution of the hot operation was small, and the long motorway distance resulted in low distance-specific emissions. Higher emissions were seen during dynamic driving for all pollutants (except NH 3 ). Nevertheless, the emissions were still relatively low (i.e., below the laboratory or on-road limits, even though not applicable to the motorway part separately). High emissions were also seen at a high ambient temperature (50 • C) for CO and NO x . The combination of high loads and exhaust gas temperatures is typically addressed by fuel enrichment for component thermal protection, which increases fuel consumption and emissions [81]. Furthermore, it could be that the 50 • C condition had not been optimized and/or calibrated. Nevertheless, for this vehicle, there were no high CO and NO x emissions on the motorway, as has been reported in some other gasoline vehicles [22,23,82]; in our tests, emissions were always below the (non-applicable) limits. The high emissions in previous studies could be also due to the small catalyst volume (i.e., high space velocity at high speeds) [83] and a rich engine operation (both result in lower conversion efficiency).

Concluding Remarks
The emissions of the vehicle of this study were low compared to other vehicles of the same emission level (Euro 6d-Temp) or older (e.g., Euro 5 or Euro 6b), as has been discussed previously. For this reason, it cannot be considered necessarily a representative vehicle of the current fleet. Nevertheless, it would not meet the proposed Euro 7 limits by CLOVE [84]. For example, the 30 mg/km (or 480 mg up to 16 km distance) would not pass the TfL cycle (around 600-800 mg) at−7 • C. Similarly, at −7 • C, the NH 3 emissions (10-20 mg/km) were higher than the proposed limit of 10 mg/km, and the CO emissions were almost double than the proposed limit of 600 mg/km, clearly indicating that future vehicles will need to be very clean.
As many researchers have shown, the majority of the emissions are emitted during the first minutes of a vehicle's ignition [85]. Although there are differences in the absolute levels between different vehicles, the cold start still remains the main contributor in most cases, and it needs to be controlled better. Thermal management methods, such as those based on burners, reformers, and electrically heated catalysts, might further decrease cold start emissions, and a degree of electrification might be necessary [27,[86][87][88][89]. However, in some cases, urban emissions of hybrid electric vehicles might be higher due to the longer time needed for the catalyst to warm up; thus, further calibration efforts are needed [32,90]. Furthermore, keeping the TWC temperature and lambda value at appropriate levels in urban stop-and-go traffic situations will be important. The difficulty of controlling cold start emissions does not mean that the rest trip is not important. It was shown that the motorway section can have an equal or even higher contribution in mass (g), even though it was relatively low as mass per distance in the vehicle used in our study. However, this is not always the case [22]. A pollutant that requires attention is NH 3 . In most cases, emissions were around 10 mg/km but, at urban cycles, they reached 35 mg/km. These levels are higher than the US limit of 6.2 mg/km and, in some cases, higher than the China limit of 20 mg/km (applicable to different cycles).
One important topic that was not addressed in this study is the catalyst's deactivation and deterioration over time [91]. The performance of a catalyst degrades over its lifetime through several mechanisms, including precious metal agglomeration, washcoat breakdown, as well as selective and non-selective poisoning. Thus, emission levels close to the useful life of a vehicle also need to be studied in the future. Finally, the effect of fuel on emissions also has to be considered.

Conclusions
A Euro 6d-Temp gasoline vehicle with TWC (three-way catalyst) and GPF (gasoline particulate filter) was tested on the road with the Type 1A on-road real driving emissions (RDE) procedure (at 17-20 • C), as well as in a laboratory with the Type 1 worldwide harmonized light vehicles test cycle (WLTC) (at 23 • C). Additional urban, motorway, dynamic, and uphill driving cycles with different payloads were conducted in the laboratory at temperatures between −30 • C and 50 • C. This is one of the few studies that has assessed a low-emitting vehicle at extreme temperatures and in extreme driving conditions. Maximum emissions at motorway cycles at 23 • C were 500 mg/km for CO, 10 mg/km for THC, 5 mg/km for NO x , and 9 mg/km for NH 3 . Maximum emissions at the congested traffic urban cycle (TfL, 8.9 km long) at 23 • C were 550 mg/km for CO, 150 mg/km for THC, and 32 mg/km for NO x and NH 3 . At −10 • C, urban emissions were around two times higher for CO, HC, and NO x , and at the same levels for NH 3 . The first 300 s contributed >75% to CO and HC and 45-90% to NO x and NH 3 of urban emissions, depending on the cycle and the ambient temperature. Compliance with the proposed future Euro 7 limits will require further emission reductions, especially at a cold start. More fundamental studies (i.e., measurement of air/fuel ratios, catalyst temperatures, etc.) are needed to better understand the behavior of vehicles in such conditions. Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/atmos12081011/s1, Table S1: Distance specific emission factors, Table S2: Fuel specific emission factors.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Acknowledgments:
The authors would like to acknowledge the technical staff of VELA for conducting the measurements.

Conflicts of Interest:
The authors declare no conflict of interest. The opinions expressed in this manuscript are those of the authors and should in no way be considered to represent an official opinion of the European Commission and the respective institutes. Mention of trade names or commercial products does not constitute endorsement or recommendation by the European Commission, the institutes and/or the authors. Figure A1 plots the coolant temperature for the TfL and BAB cycles for various ambient temperatures. The coolant temperature had an increasing trend throughout the TfL cycle at all ambient temperatures, except 50 • C, where it stabilized after 700 s. A temperature of 70 • C was reached after 1900-2100 s, when the ambient temperatures were −7 • C and −10 • C, respectively, and after 1200 s, when the ambient temperature was 5 • C. The coolant temperature remained relatively constant at 105 • C during the BAB cycles (OBD data at 23 • C were not logged). These results indicate that the engine did not reach thermal stability during the TfL cycle at all ambient temperatures (except 50 • C). At 1700 s at 50 • C, the coolant temperature slightly dropped, indicating that an active thermal management strategy (engine measures to reduce the combustion temperature) might have been implemented (e.g., fuel enrichment, late intake valve opening), which could influence (engine-out) emissions. Figure A2 plots the coolant temperature for the RDE short cycles for various ambient temperatures. The coolant temperature stabilized around 90 • C at 50 • C and −30 • C after 200 s and 2400 s, respectively. It stabilized at 105 • C at 5 • C and −10 • C after 1500 s and 2000 s, respectively.   Figure A3 plots the cumulative emissions of various pollutants for the RDE short cycle. Figure A4 presents the coolant temperatures during the first 15 min of various urban cycles at −7 • C and/or −10 • C. The coolant temperature reached approximately 100 • C after 500-600 s for the high load uphill cycles and after 800 s for the dynamic boundary RDE. The RDE short cycle reached 70 • C and the TfL reached 40 • C after 900 s.