High-resolution spatial distribution maps of road transport exhaust emissions in Chile, 1990 – 2020.

. This description paper presents a detailed and consistent estimate and analysis of exhaust pollutant emissions generated by 15 Chile´s road transport activity for the period 1990-2020. The complete database for the period 1990-2020 is available at doi: http://dx.doi.org/10.17632/z69m8xm843.2 (Osses et al., 2021). Emissions are provided at high-spatial resolution (0.01º x 0.01º) over continental Chile from 18.5 S to 53.2 S, including local pollutants (CO, VOC, NOx, PM 2.5 ), black carbon (BC) and greenhouse gases (CO 2 , CH 4 ). The methodology considers 70 vehicle types, based on ten vehicle categories, subdivided into two fuel types and seven emission standards. Vehicle activity was calculated based on official databases of vehicle records 20 and vehicle flow counts. Fuel consumption was calculated based on vehicle activity and contrasted with fuel sales, to calibrate the initial dataset. Emission factors come mainly from COPERT 5, adapted to local conditions in the 15 political regions of Chile, based on emission standards and fuel quality. While vehicle fleet has grown fivefold between 1990 and 2020, CO 2 emissions have followed this trend at a lower rate and emissions of air local pollutants have decreased, due to stricter abatement technologies, better fuel quality and enforcement of emission standards.


Introduction
Building and updating emission inventories provides key information for designing and evaluating public policies concerning 30 topics relevant for the inhabitants of cities' quality of life, the environment and for mitigation of climate change (Kuenen et al., 2014;Creutzig et al, 2015). In international and national experiences, the construction of reliable emission inventories for road transport has been a bottleneck in mapping the emissions in cities (Zheng et al., 2014). In general, these difficulties occur due to two main reasons: the lack of disaggregated data at a level to construct detailed inventories, and the many variables to consider when modelling emissions, increasing the uncertainty in the estimation of total emissions (Bond et al., 2004;Tolvett 35 et al., 2016).
Latin America has an urbanization rate of more than 80% and cities suffer changes in local climate and air pollution, imposing big challenges to cope with (Henríquez and Romero, 2019;Hardoy and Romero-Lankao, 2011). Transport in cities is a major source of air pollution and emissions of greenhouse gases (GHG) (Huneeus et al 2020a). Reliable inventories are needed to 40 assess policy measures for air quality and climate change. In the case of Santiago, Chile's capital, there are good examples of the use of local data to analyse the impact of emissions on air pollution (Mazzeo et al., 2018), health benefits of policy scenarios (Mena-Carrasco et al., 2012), and for retrospective evaluation of the evolution of mobility and air quality, relating them to policy measures . by political region, which were used to calculate fuel consumption by vehicle category and, subsequently, exhaust emissions, as summarized in the methodological diagram ( Fig. 1) and explained in this section. The resultant emission database was 95 spatially distributed at 0.01x0.01 degrees of latitude and longitude resolution. Vehicle fleet composition was based on official government data on annual registration of in-use vehicles, i.e., the vehicles that each year pay their circulation permit after having approved the periodic technical inspection. The National Institute of Statistics (INE, https://www.ine.cl/) provides annual reports of the total number of vehicles with circulation permit per political region. Vehicle categories reported are light passenger, commercial and taxi vehicles; 12-and 18-meter buses; light, medium and heavy-duty trucks; and two-wheeled motorized vehicles. Since the emission factors and fuel consumption of the distinct 105 types of vehicles are being modelled using the Computer Programme to calculate Emissions from Road Transport (COPERT) version 5 developed by the European Consortium of Transport Impact (EMISIA, https://www.emisia.com) (Ntziachristos et al., 2009), the equivalence between Chilean INE categories and COPERT is shown in Table 1. Each of these categories was subdivided to distinguish the type of fuel used (gasoline or diesel) using the most recent information from the Ministry of Transport and Telecommunications (MTT) and the secretariat for transportation (SECTRA, 2014)   The combination of categories, fuels and emission standards generates a total of 70 types of vehicles for the emission analysis, distributed over political regions and distinguishing between urban and interurban activity. The distribution of vehicles into urban and interurban activity per region was based on a proportional regional distribution according to SECTRA (2010). 125

Calculation of vehicle activity and fuel consumption
Activity level is expressed as VKT (vehicle kilometre travelled) calculated as the sum of the vehicles in each type per kilometres driven (Eq. 1).
where Ni,j,k is the number of vehicles of type i in region j and road class k (urban or interurban); KMi,j,k are the kilometres 130 travelled per year by vehicles type i, in region j and road class k.
The kilometres travelled by each type of vehicle used in equation 1 are shown in Table 3. They correspond to estimations by SECTRA (2014) andMAPS (2013) for the first level of vehicle aggregation (main vehicle categories on Table 1).

135
where is the region to which the vehicle belongs; represents the vehicle type (passenger, commercial, bus, truck, 145 motorcycle); represents the share of the subcategory for the distinct vehicle types: for passenger and commercial vehicles the subcategories are diesel and gasoline, for buses are rigid, articulated and intercity and for trucks are light, medium and heavy duty; represents the share of vehicles that drive on urban areas or interurban roads; is the share of vehicles according ' is the percentage of the distinct subcategories, ( is the share of traffic counts depending if the vehicle is driven in urban or rural areas and ) is the percentage of vehicles with distinct emission control technologies. Thus, Equation 2 is used to calculate the total fuel consumption for each category presented as sub-indices.

8
TFC was compared to real fuel sales for each region. The Electricity and Fuel Superintendence (SEC, www.sec.cl) provides information on sales of diesel and gasoline for the transportation sector, by political region. The calculated TFC (see Equation   2) was compared to the data given by SEC and then a correction factor to the total number of registered vehicles in each region is applied to make these two fuel consumptions equal, correcting for those vehicles that are registered but do not contribute to actual driving activity. Thus, the number of active vehicles in a region was inferred and adjusted accordingly. A comparison 160 between official figures of national fuel sales (SEC) and estimated TFC, for gasoline and diesel at a national level, is shown in Figure 2. In general, estimates of fuel consumption are lower than fuel sales, which means the number of registered vehicles generates lower activity than reality or specific consumption factors for each vehicle technology are lower than the real driving conditions in Chile. These differences are addressed increasing the number of vehicles according to their technology, matching official fuel sales figures. 170
With the calculated mobility demand, emissions of pollutants based on vehicle-kilometres travelled by the different vehicular 180 categories can be estimated through Eq. (3) (Zheng et al., 2014): where the sub-index n represents the pollutant: CO2, CO, NOX, PM2.5, BC, VOC and CH4. EF is the emission factor for the pollutant n [ . &' ] for each vehicle. 185 CO2, CO, NOX, PM2.5, VOC emission factors are being modelled using COPERT 5 methodology (Ntziachristos et al., 2009) based on the average speed in the driving cycles given by previous studies in all regions of Chile (Osses et al., 2014). CH4 emission factors used are from previous reports (Ntziachristos et al., 2007;USEPA, 2018). However, COPERT 5 does not model black carbon emission factors due to the difficulty of the classification of this type of aerosol (Bond et al., 2013). 190 Nonetheless, there have been studies to determine black carbon fractions in particulate matter, distinguishing between vehicle technology and fuel type, considering that elemental carbon and black carbon fractions are equivalent (Bond et al., 2004;Chow et al., 2010;Minjares et al., 2014;Ntziachristos et al., 2007). These fraction values can be used to obtain BC emission factors as follows: where *+ is the emission factor of the total particulate matter in the exhaust, (.-is the mass fraction of particles that have an aerodynamic diameter of 2.5 or less and ./ is the fraction of black carbon in these particles. The mass fraction of fine particles used is 0.9 since in a generic and ideal particle distribution, between 80 and 95% of the total mass of the particles are concentrated in this range (Payri and Desantes, 2011). Table 4 shows the black carbon fractions used, distinguishing vehicle category, motorization and vehicle technology. Motorcycle BC emission factors used were taken from the literature (Cai et al., 200 2013). Roman number notation applies for heavy duty vehicles and Arabic numbers apply to light commercial and passenger vehicles.

205
COPERT V considers correction of emission factors by vehicle age for light vehicle categories EURO 3 & 4 and for VOC, CO, NOx. These corrections were also applied.

Spatial disaggregation 210
The spatial distribution of transport emissions per political region consists of allocating Gg of emissions per year to each cell in a grid, with cells of 0.01x0.01 degrees of latitude and longitude covering the fifteen regions in the country. The regions correspond to the political administrative division of the territory. The distribution depends on the types of roads in each cell, the vehicle flow, and the presence of urban population.

215
The identification of roads in each cell was based on road network maps, available from the official roads database for Chile´s and the official regional limits (BCN, 2020). The information on Chile´s road network was complemented with data from OpenStreetMap (OSM, 2020). It covers a total of 77.800 km of rural and urban roads. Each road on the network was classified into a hierarchy comprising freeways, arterials, collectors, and local roads. The estimation of vehicle flow on each type of road resulted from applying a road weight factor, based on toll barrier vehicle counts at interurban roads (MOP, 2020) and origin-220 destiny surveys in urban roads. Average weight factors are 54% for freeways, 23% on arterials, 16% for collectors and 7% on local roads. The road weight factors vary by region, urban and interurban areas, and cities in a region, and are provided by the Transport Secretariat, SECTRA (Osses et al., 2010).
Emissions were distributed over the road network using QGIS opensource software. Urban emissions were also distributed 225 among the cities of each region according to their population (INE, 2017). QGIS allocates emissions to cells based on the type of roads, with their weight factor, and the presence of cities. Therefore, emissions for each cell depend on the roads and the presence of urban population. The Transport Secretariat, SECTRA, provides the proportion of urban and interurban roads per region (Osses et al., 2010) and the urban areas of each region can be associated to cities with population over 5000 inhabitants.
The number of kilometres in each cell is proportional to the annual emission for each cell in the grid, and the sum of emissions 230 in all cells coincide with the total emission assigned to each region of the country.

Evolution of fuel quality, vehicle technology and emission factors
Based on the information given in sections 2.1 and 2.2, the number of vehicles and their technological evolution has been determined, according to European emission standards (Pre EURO, EURO 1-6 for light duty vehicles, EURO I-VI for trucks 235 and buses). The enforcement of stricter emission standards along the country has been sustained by permanent national fuel quality improvements. The reduction of sulphur in fuels have been progressive since 1990 to 2004 from 5000 ppm to 50 ppm of sulphur in diesel and from 1000 ppm to 30 ppm in gasoline. In April 2001, the elimination of leaded gasoline was made effective, which allowed national enforcement of three-way catalytic converters, reducing the levels of CO, VOC, NOx and also particulate matter from motor vehicles (Moreno et al., 2010).  (Table 2), are expected to keep decreasing the exhaust emissions of local pollutants from on-road transportation, even considering the permanent increase of mobility. The evolution 245 of emission standards and number of vehicles for the whole country is shown in Figures 3 and 4, adding up specific regional information from 1990 until 2020.   (Pre EURO) and EURO 1 standards is also observed, ending in 2020 with a fleet mixed between EURO 3 and EURO 5. By 255 2020 there are already a few EURO 6 vehicles, but they cannot be distinguished in the graphs.   Knowing the different vehicle technologies in Chile and their equivalence with the European categories (Table 1), it is possible 275 to obtain the emission factors from the values reported by the COPERT model (Ntziachristos et al., 2009). For doing this, different activity parameters are considered, where the most relevant are the average speed of displacement and load level. Table 6 shows the result of this assignment, considering urban speeds for light vehicles, buses and motorcycles, and interurban speeds for trucks. The vehicle categories correspond to those indicated in Table 5.  In total, 70 vehicle categories are generated, which are doubled to 140 types of emission when considering urban and interurban travelling speeds. All these emission types apply to the different pollutants, which are shown in Table 6. In general, all emission factors decrease as the level of the EURO standard is increased. CO2 emissions for gasoline vehicles are higher than diesel 290 vehicles, this compound being the one with the least reductions since all technologies burn fossil fuels. Diesel vehicles contribute most of the emissions of PM2.5, BC and NOx, but with important reductions when going from EURO 4 / IV to EURO 5 / V or EURO 6 / VI. It is interesting to note the high emission factors of PM2.5 and CO from motorcycles, especially considering their recent increase in the Chilean vehicle fleet.

Annual emission trends at a national level 295
Using the activity levels and emission factors previously described, total emissions are calculated by pollutant, vehicle type, and country region. Table 6 shows a summary of the total annual emissions, with the variation percentages between 1990 and 2020. CO2 has an increase of 207.7%, compared to 309% in mobility growth (VKT) for the same period. CO2 official values for on-road transportation have been reported by Chile from 2010 until 2018 (MMA, 2018), differing by ±1.4% with values shown in Table 6 during those years. 300 Table 6. Total annual exhaust emissions produced by on-road transportation in Chile, 1990Chile, -2020 Year  Unlike CO2, the rest of the local pollutants included in Table 6 are decoupled from the growth in mobility, reducing their 305 contribution significantly thanks to technological improvements. NOx is the one with the least reduction compared with economic growth, with 20.4% of emissions in 2020 compared to 1990. Emissions of PM2.5 and BC go up for the first 20 years of analysis, starting to decrease after 2010, where the reduction ratio of PM2.5 is greater than BC. CO and VOCs, mainly associated with gasoline engines, show significant reductions, mainly due to the massive incorporation of three-way catalytic converters required for vehicles complying with EURO standards. 310  Table 7 and Figure 5 show annual emission trends for six different compounds, divided by vehicle type. In the case of CO2 and NOx, all types of vehicles have relevant contributions ( Fig. 5a and 5b); diesel vehicles are responsible for the majority of 315 PM2.5 and BC emissions ( Fig. 5c and 5d); and gasoline cars dominate CO and VOC emissions ( Fig. 5e and 5f). From 2008 onwards, BC emissions are not mitigated with the same rate as PM2.5 does, the BC reduction being less effective, 340 which can contribute to local health problems and negative effects on local climate change (Bond et al., 2013;WHO Regional Office for and Europe, 2012). This can be explained given that the BC / PM2.5 fraction for light vehicles decreases significantly when going from EURO 4 to EURO 5, but this is not the case for heavy vehicles, which have this significant decrease later, between EURO V and EURO VI (Table 4). Trucks and buses are the main contributors to the total emissions of BC, and these two categories are the latest to be required with EURO VI. According to the information given in Table 4, the BC fraction in 345 PM2.5 emissions for heavy duty trucks is 75% in EURO V, which is not a considerable reduction compared to other vehicle types.
Finally, all the emission curves show a drop in the 1999-2004 period, which is explained by the impact that the Asian financial crisis had on Chile, significantly affecting the sale of motor vehicles and their activity. 350

Spatial disaggregation
The complete database for the period 1990-2020 is available at doi: http://dx.doi.org/10.17632/z69m8xm843.2. This inventory is part of the first gridded national inventory of anthropogenic emission for Chile of criteria pollutants as well as GHG (hereafter INEMA from Spanish Inventario Nacional de EMisiones Antropogénicas), presented by Alamos et al (2022).
INEMA comprises emissions for vehicular, industrial, energy, mining and residential sectors for the period 2015-2017 in Chile. 355 The spatial disaggregation of emissions at the national level shows the high concentration of emissions in urban areas and main roads in the country. Figure 6 shows the fraction of PM2.5 emissions for the year 2020 over the cells of 0.01x0.01 degrees of latitude and longitude, which is equivalent to approximately 1.11x1.11 kilometres. It is difficult to clearly identify these emissions on the complete map of Chile, due to its shape (Fig. 6a). However, when zooming in on each region, the populated 360 areas with high emissions on the road network appear. Fig. 6b shows the city of Antofagasta, which is approximately 22 kilometres long and has a population of 388 thousand inhabitants, which concentrates most of the on-road vehicle activity. The images in Figure 7 show the fraction of NOx emissions for the year 2020 in six major cities of Chile, from north to south.
The emission grid is superimposed with the road network, green areas, and uninhabited areas, obtaining a good matching between them. In each city, the areas of greatest activity are coloured with warmer tones, indicating greater NOx emissions, 375 decreasing towards colder tones for a lower fraction of these emissions.

Comparison with previous results
A direct comparison of emissions from this study with other emissions estimates was performed to reflect the differences in estimation approaches between local (bottom-up) and global (top-down) models, as well as the sensitivity to different 380 assumptions in the estimates. Figures 8, 9 and 10 show the comparison among local estimates from this work -INEMA, the National Emissions Inventory -INGEI (MMA, 2020) and an estimate using the LEAP model (Kuylenstierna et al., 2020); and global estimates by the EDGAR V5.0 global model (Janssens-Maenhout et al., 2017) -EDGAR, the CAMS-GLOB-ANT v.
4.2 dataset -CAMS (Granier et al., 2019), and the CEDS dataset -CEDS (McDuffie et al., 2020;Smith et al., 2015), for CO2, CH4, PM, BC, CO, and NOx from 1990 to 2020, according to the pollutants available in each estimate. It is worth mentioning 385 that EDGAR, CAMS and CEDS are not independent. For historic years CAMS is mostly based on EDGAR but extrapolated to more recent years, using other information such as trends from CEDS.  Furthermore, after 1999, these global inventories show a consistent increasing trend. Such trend that was not followed by local estimates, which show a stabilization between 1997 and 2005, and a rather consistent decrease since 2007. As a result, EDGAR and CAMS PM (BC) emissions from 1999 to 2015 were between 85% (87%) and 315% (208%) higher than those from the local inventory. On the other hand, CEDS estimates for BC were even higher than EDGAR and CAMS estimates for the whole 410 period, although they followed similar trends between 2000 and 2015. CEDS/INEMA BC emission ratios range from 2.76 to 5.69, suggesting that BC emission factors in the CEDS dataset are significantly higher than those used in this work. Since this work's emission factors are based on the COPERT model and the actual vehicle technology distribution, higher PM and BC emission factors used in EDGAR and CEDS imply assumptions of an older fleet in global inventories. Standards for diesel vehicle emissions and sulphur fuel content have been greatly improved in Chile since 2000, so EDGAR, CAMS and CEDS 415 emissions and increasing trends for PM and BC are likely overestimated.

435
Finally, the CO/NOx ratio and its trends are shown in Figure 11, which not only includes Chile but also a comparison with European and two other countries in the Latin America and the Caribbean (LAC) region between 1970 and 2020. The CO/NOx ratio was much higher in the global inventories than in the local inventories, with a mean difference of 209% [90% -457%] between EDGAR and INEMA estimates for Chile, and shows a decreasing trend in both, with more fluctuations in the global inventory. The differences in emissions and trends for CO and NOx suggest that global emission inventories use emission 440 factors that correspond to technologies older than those that have been and are currently used in Chile. Considering the differences between EDGAR data and this study's results for Chile, trends in CO/NOx ratio for other European and LAC countries from EDGAR were included. A big difference appears between these two groups of fleets, the CO/NOx ratio being much higher for LAC selected countries. In other words, according to EDGAR figures, LAC CO/NO ratios reach European values 40 years later (1970 versus 2012), which seems inaccurate according to local estimates. Chile's CO/NOx ratios are in 445 the same range as those found in European countries, which is supported by the fleet renewal shown in Figures 3 and 4. Most of the Chilean fleet consists of Euro II/2 and Euro III/3 vehicles, which have much lower CO/NOx ratios that pre-Euro ones.
Our analysis for Chile suggests that a careful analysis of national versus global estimates and/or emission factors for road transport emissions is needed for other LAC countries as well. External data obtained from models EDGAR/CEDS/LEAP

Conclusions 455
This paper describes an original dataset for transport emission in Chile between 1990 and 2020, spatially distributed at 0.01º x 0.01º. The dataset is based on annual reports from governmental agencies, and estimates the evolution of air pollutants (CO, VOC, NOx, PM2.5), greenhouse gases (CO2, CH4) and black carbon (BC). Results were contrasted with EDGAR, CAMS and CEDS datasets.

460
The analysis shows a significant growth of the vehicle fleet coupled with increasing CO2 emissions, which agree with the national inventory of GHG. Air pollutants show different patterns, with a general decreasing trend which coincides with pollution control measures. Data shows a clear relationship between these emissions and the introduction of better fuel quality, due to reduction of sulphur content, and enforcement of technological improvements. These policy measures included regulation of emission standards for new vehicles into the fleet, mandatory periodic technical inspection for in-use vehicles, 465 as well as effective procedures for regulation enforcement.
The comparison with EDGAR, CAMS, CEDS, and locally estimated datasets shows agreement in CO2 estimations and striking differences for local compounds, with global estimates consistently higher. This disagreement is likely due to differences in assumptions of vehicle technologies characterizing the fleet and quality of the fuel used. In the case of PM and BC trends 470 between EDGAR and this transport dataset diverge from 1998, for CO, NOx and CH4 since 2006 -2008. Results suggest that global emission inventories use emissions factors that do not coincide with the technologies of the vehicle fleet. EDGAR assumes a 40-year delay in technological update for Latin-American vehicle fleets compared to European ones, which is inaccurate for the case of Chile according to the dataset presented in this paper.

475
Every dataset has limitations and this is not an exception, INEMA does not include cold start emissions, neither consider a calibration of fuel consumption according to vehicle age. The use of international emission factors is a second best compared to using locally measured emission factors and COPERT does not cover aging for all vehicle categories in the dataset. The impact of COVID-19 is not considered in 2020, but other studies have addressed these effects on urban emissions in Santiago (2020). However, these limitations should not significantly change the results of the paper since the database provided is more 480 accurate and extended than the existing ones, and the comparative analysis with external datasets show differences that need attention.
This paper illustrates the potential of local datasets for policy ex-post impact assessment. It also reinforced the value of available official raw data, produced with transparent methods and on a regular basis, as well as the production of national 485 inventories. Further work could build on the dataset presented in this paper to produce projections and scenarios for future policy making. Work should be done on the construction of local emission factors, this is the only information of the modelling that is not produced locally, real emissions campaigns of a sample of the fleet could strengthen the results of this analysis.

Data availability 490
This dataset contains annual exhaust emission inventories of CO, VOC, NOx, PM2.5, CO2, CH4 and BC from on-road transportation in Chile, for the period 1990-2020. The data is presented as NetCDF4 files, in Gg/y per cell for each specie and year, gridded with a spatial resolution of 0.01° x 0.01° covering the domain 66°-75° W and 17°-56° S. It can be accessed through the open access data repository http://dx.doi.org/10.17632/z69m8xm843.2, under a CC-BY 4 license (Osses et al., 2021).

Competing interests
Author MO is a guest member of the editorial board of the journal. 505