Computational model integration for high-resolution spatial emission simulation: a tool for urban transport management in resilient cities

Air pollution represents a critical environmental risk to public health, particularly that derived from the combustion of fossil fuels, which may account for millions of premature deaths annually worldwide. In intermediate Latin American cities, the sustained growth of the vehicle fleet and fragmented urban planning exacerbate this situation. High-spatial-resolution diagnostics of vehicle emissions could facilitate sustainable mobility policies. This research proposes an integrated methodological model to optimize the spatial disaggregation of bottom-up vehicle inventories in contexts of limited information for small and medium-sized cities. The proposal links three sequential computational models using the intermediate city of Tandil, Argentina, as a case study: 1) base emission estimation using the International Vehicle Emissions (IVE) model; 2) road network characterization through GIS based on demographic, commercial, and educational densities; and 3) adaptation of the open-source DROVE algorithm in R. en road categories were identified to integrate the functional heterogeneity of the urban system. A high-resolution spatial disaggregation for criteria pollutants —specifically PM10 —was analyzed. By incorporating urban variables, simulation precision in DROVE was significantly improved, with results further enhanced as the spatial resolution of the grid increased. This integration of sociodemographic factors allowed for the classification of the road network according to urban behavior patterns, successfully identifying critical points and corridors of pollutant loads that would otherwise remain attenuated. The proposed methodological sequence accounts for the heterogeneity of the data reflecting the urban system, disaggregating emissions from inventories by leveraging local spatial databases. This reduces dependence on vehicle counts and traffic models, which are often the most significant constraints in resource-constrained contexts. The contribution to mobility management and SDG 11 (Sustainable Cities and Communities) is achieved by focusing on less-studied urban scales with high global representativeness, providing an accessible and efficient tool for studying and understanding vehicle emissions.

Currently, atmospheric pollution constitutes one of the main environmental risks to health. It is estimated that air pollution caused between 4.2 and 4.6 million premature deaths for the year 2021 (WHO, 2024; Health Effects Institute, 2024). Recent evidence indicates that a significant proportion of the mortality attributable to PM 2.5 is specifically linked to emissions derived from the combustion of fossil fuels, which could explain millions of annual premature deaths globally. Furthermore, it has been observed that the relationship between exposure and risk is non-linear, with relevant effects even at low concentrations. Given that these emissions are largely regulable, this component represents a critical point for action in public health and environmental policy (Vohra et al., 2021). In this context of high sanitary burden and identifiable emission sources, it is especially relevant to consider that exposure and its effects are not distributed in a homogeneous manner in the population. There are vulnerable groups with pre-existing diseases and unfavorable socioeconomic conditions, for whom there is no safe level of exposure nor a threshold below which adverse health effects do not occur (Elkama et al., 2021). Nonetheless, various regulatory frameworks establish guideline values or maximum permissible limits for the concentration of pollutants, standards that serve as thresholds for operational references based on acceptable risk levels (European Parliament and Council of the European Union, 2024; WHO, 2021).

In intermediate cities of Latin America, the sustained growth of the vehicle fleet, in combination with fragmented urban planning processes, constitutes one of the main sources of environmental degradation (Nieuwenhuijsen, 2020). Argentina is one of the most urbanized nations in the world, with 92% of its population living in urban areas and a projected growth of 21% between the years 2020 and 2050 (United Nations, 2019). However, current public policy still does not contemplate a progressive abandonment of fossil fuels for transport, which perpetuates the dependence on internal combustion technologies (Climate Transparency, 2021). Within the framework of the Sustainable Development Goals (SDGs), advancing toward SDG 11, Sustainable Cities and Communities, requires overcoming the technical deficiencies that condition decision-making.

In many intermediate cities of the region, the massive adoption of zero-emission technologies (e.g., electric mobility) will not occur in the short term (Calatayud et al., 2023); therefore, it is indispensable to manage the impacts of transport through evidence-based decisions. Nonetheless, unlike large metropolises, these urban areas frequently lack baseline information and, as a consequence, the scarcity of diagnostics of high spatial and temporal resolution limits environmental governance and hinders the implementation of effective sustainable mobility policies.

Emissions modeling, as well as traffic modeling, has rapidly developed in recent years. Emissions models are created from chassis dynamometer tests where experimental emissions data are collected under laboratory conditions, such as Federal Test Procedure - FTP75- and Highway Fuel Economy Test - HWFET- (Tapia et al., 2017; Code of Federal Regulation, 2024). Technological developments in measurement equipment allowed for the collection of emissions data in real-world road driving situations using portable emissions measurement systems (PEMS) (Zhai et al., 2020), which were then used to develop more precise emission factors and models through processing (Mądziel et al., 2022; Wen et al., 2021). Within the context of microscale models (Mądziel, 2023), these are important for obtaining vehicle speed and acceleration profiles in situations involving terrain slope, driving patterns, engine load, and traffic changes at different times of the day, seeking to adjust the models to real trajectories.

The evaluation of the impact of transport on public health and the environment requires a sequential modeling chain whose stages are strictly interdependent. According to the process approach of the International Organization for Standardization 9001 (ISO) (2015), articulating these interrelationships and interdependencies is essential to ensure overall performance and the reliability of the results. In this framework, the articulation begins with the development of a mobile source emissions inventory, which provides the input data for atmospheric dispersion models and, consequently, for population exposure studies. Although the application of bottom-up approaches—such as the Vehicle Emissions Inventory (IVE) model—allows for an accurate estimation of total emissions from representative vehicle traffic counts, the main challenge in intermediate cities arises in the subsequent stage: the detailed spatial disaggregation of that inventory over the entire road network (International Sustainable Systems Research Center ISSRC, 2008).

Achieving correct spatial allocation and preventing concentrations from being distributed by assuming homogeneous behavior requires coupling the inventory with exhaustive vehicle flow models (link-based models) or having continuous traffic counts for each section of the city. However, once the initial stage of the inventory is passed, growing urban areas rarely have this massive information. It is at this methodological moment that it becomes imperative to adapt complementary tools that facilitate the transition between the global inventory and the territory, overcoming the technical barrier imposed by the lack of traffic data.

There is great potential in leveraging databases for demographic and urban management. These integrate socioeconomic information from the census with road networks and land use coverages processed through Geographic Information Systems (GIS), making it possible to quantify characteristics at the micro-level. Census tract analysis reveals detailed patterns of spatial structure that generic density measures often omit. The incorporation of these indicators favors a more precise calibration of simulation models, making them more sensitive to the urban structure (Srinivasan, 2001). In this context, the larger volume and level of disaggregation of the data contribute to the development of models with better spatial resolution (Jing et al., 2016). In turn, advances in cartographic and geospatial technologies point toward high-resolution emissions models that recover specific characteristics of each location (Batterman et al., 2015), which is particularly relevant for traffic and emissions simulations in urban areas.

To respond to this challenge, this work proposes an integrated methodological framework that optimizes the spatial disaggregation of bottom-up vehicle inventories in contexts of limited information. The strategy combines the base estimations obtained through the IVE model with the use of GIS and the adaptation of the open-source Algorithm for Spatial and Temporal Disaggregation of On-road Vehicle Emission Inventories (DROVE).

The objective of this research is to build an integrated computational model that allows for a more detailed understanding of the disaggregation of the vehicle emissions inventory on an urban scale, considering the city as a system that provides key information and variables for the construction of this model. The work addresses criteria pollutants (PM 10, CO, NO x, SO x) with models applicable in intermediate localities with scarce traffic information resources, and uses Tandil (Argentina) as a model city. This seeks to overcome the inherent limitations of aggregated inventories, providing precise inputs for air quality models and equipping decision-makers with a robust and accessible territorial management tool to promote environmental resilience and sustainable urban mobility.

The integration of computational models with methodological approaches that allow us to explain vehicle emissions and spatial concentrations at an urban scale was carried out in the city of Tandil (Figure 1). The proposal links three methodological models with their respective computational designs; the workflow sequence is described in the following items.

Study area and data collection

The locality of Tandil covers an approximate area of 10,155.13 ha and represents an example of an intermediate city in Latin America and the world (United Nations, 2025); it is located in the southwest of the Province of Buenos Aires (Figure 1). According to the National Census conducted in 2022, the Tandil district has 145,575 inhabitants and the city has 138,549 (Instituto Nacional de Estadística y Censos (INDEC), 2023)

The area presents a humid temperate climate with hilly influence, characterized by the convergence of air masses of different origins and predominant winds from the north and south sectors (Ruiz de Galarreta et al., 2007; Lavornia, 2015). Its urban expansion is conditioned toward the southwest by the Tandilia hill system, which acts as a physical limit to building growth.

From a mobility perspective, Tandil’s road system is structured as a hierarchical network of highways, avenues, and streets that articulate urban and regional connectivity. Regional roads (such as National Route No. 226 and Provincial Routes No. 30, 74, and 80) ensure interurban integration, while main and secondary avenues organize the city’s internal flows. Local streets and service roads complete the network, facilitating access at the neighborhood and parcel scale. This system is regulated by planning criteria that establish functional hierarchies and minimum dimensions, aimed at guaranteeing connectivity and the ordering of urban growth (Municipalidad de Tandil, 2005).

The private vehicle fleet consists of a total of 54,305 vehicles that routinely travel through the local urban area. Regarding the dynamics of the vehicle fleet, the municipality evidenced a sustained growth during 2025 in both new (0 km) vehicles and used vehicle transfers. The new vehicle segment represented a year-on-year growth of 44.8%, a figure similar to those observed at the provincial (+48.8%) and national (+47.7%) levels. Meanwhile, in the used vehicle market, there was a growth of 5.1%, comparable to those registered in the province (+6.6%) (Dirección de Estadística Local, 2025).

Estimation of baseline emissions. bottom-up inventory

According to Mądziel (2023), the IVE model for estimating vehicle emissions into the atmosphere can be classified as a microscale estimation model. In these models, vehicle acceleration and speed are the most important simulation parameters. The IVE model has been developed as a joint effort of the University of California at Riverside, College of Engineering–Center for Environmental Research and Technology (CE-CERT), Global Sustainable Systems Research (GSSR), and the International Sustainable Systems Research Center (ISSRC). Its purpose is to estimate vehicle emissions in different geographical areas, especially in developing countries, providing an effective, economical, and reliable methodology for processing such data (Amaya, 2006; Viteri et al., 2023).

IVE is a model designed to efficiently collect vehicle data for the precise estimation of vehicle emissions in urban areas. Emissions from on-road mobile sources vary considerably according to three factors: 1) vehicle type, 2) driving pattern, and 3) local climate and terrain conditions (locations input data). Vehicle type refers to the source technology, that is, the type of fuel used, air/fuel ratio, engine size, emission control technology, and cumulative use (total kilometers traveled) which determines the age of the vehicle. The driving pattern or driving behavior is described by a speed profile measured in real local driving situations, the number and distribution of vehicle starts, and the daily kilometers traveled. Finally, locations input data includes fuel composition, ambient temperature, ambient humidity, and operating altitude (above sea level) among other local conditions that affect vehicle emissions; here, the differences between road hierarchies are represented (Figure 2).

As output data, IVE provides total daily and hourly emissions for each inventoried vehicle category and each pollutant. It should be noted that it does not present a spatial distribution of the emissions accounted for by the inventory, and that the estimated concentrations are constructed for a registered fleet and an average annual mileage, without losing sight of the aforementioned adjustments to adapt the emission factors used by the model.

To carry out the emissions inventory in Tandil under a bottom-up approach, primary data sources were used. Vehicle counting campaigns were conducted on arteries with different road hierarchies (arterial, highway, residential) in three distinct socioeconomic sectors of the locality (low, medium, high); for this purpose, a Bosch Intelligent Traffic Systems camera was used (model NBE-7604-AL-OC, DINION inteox 7100i IR). The counting sessions took place on weekdays between 7:00 a.m. and 9:00 p.m. (21:00), taking one measurement in each artery hierarchy. The driving pattern measurement campaigns were carried out in seven circuits: one arterial and one residential for each stratum, and an additional one to survey National Route No. 226, which crosses the city; the same criteria for days and times were followed on the arteries. To record instantaneous speed, smartphones with the GPSLogger application were used as instruments; thus, coordinates for geopositioning were collected at regular intervals (data per second). Another necessary data point for the inventory consisted of the count of daily vehicle starts during the weekdays of a typical day (frequency of starts and stops in 24 h); these data were collected through a survey, reaching a total number of 30.

Secondary data corresponding to the private vehicle fleet were obtained for the year 2024. Information provided by the Regulatory Body for Vehicle Technical Inspection of the Province of Buenos Aires (Ente Regulador de la Verificación Técnica Vehicular) was analyzed to obtain the technologies and the total number of the active or circulating fleet; furthermore, this data point was adjusted with information from the local Automotive Registries.

Meteorological data to feed the emissions model were provided by the National Meteorological Service, Aero Tandil Station for the period from 1-1–2024 to 12–31-2024. Hourly meteorological variables of relative humidity and temperature were used.

GIS-based characterization and typological classification of the road network

The analysis relied on Geographic Information Systems (GIS) for the extraction, processing, and normalization of spatial variables and was developed in two complementary stages. In the first stage, the dimensions and variables to be analyzed for each network segment were constructed from vector layers and spatial operations: road hierarchy was assigned based on the municipal functional classification and OpenStreetMap (Figure 3A); commercial density was calculated using linear buffers normalized per 100 m of segment (Figure 3B); demographic density was determined through the proportional redistribution of the census tract population (Figure 3C) based on road length and block structure; and the density of educational establishments was calculated by counting facilities within a 500 m buffer and standardizing per 100 m of segment (Figure 3D). The analysis focused on the comprehensive classification of the urban road network of Tandil based on the combination of four dimensions -one with categorical values and three with quantitative values. All quantitative variables were thus expressed as standardized physical densities per 100 m of road segment to ensure direct comparability and real-world urban interpretation.

In the second stage, a typological classification of the segments was performed using k-means non-hierarchical clustering techniques (Jain, 2010). Since road hierarchies are qualitative variables, they were maintained as fixed categories and were not included in the construction of the attribute space; instead, the algorithm was applied within each hierarchy to identify internal typologies based on the three quantitative densities. The variables were previously standardized; the optimal number of clusters was evaluated using validation criteria (elbow method and Pseudo F-statistic), and the results were spatially cross-referenced in the GIS to verify consistency with the urban structure and available traffic counts. The methodological objective was to construct a multivariable typology of road segments capable of capturing spatial patterns associated with urban functions, intensity of use, and the socioeconomic dynamics present in the city.

Adaptation of the DROVE algorithm for spatial disaggregation

DROVE is a computational algorithm for the spatial and temporal disaggregation of on-road vehicle emissions developed by González et al. (2020). It consists of open-source code developed in R that provides the distribution of emission flows based on the estimation of disaggregation factors (DF). These DFs require a grid cell domain with a specific spatial resolution (ranging from several km to a few meters) that can be constructed using GIS tools such as QGIS or ArcGIS. The algorithm offers three possible approaches with different levels of local information detail to perform the spatial disaggregation of emissions: length of road segments (LRS) for Option 1, LRS + road type for Option 2, and LRS + traffic flows for Option 3. Additionally, it contains functions for the daily and hourly temporal disaggregation of emissions.

The architecture of the algorithm is based on 2 R scripts. The first is “UserInterface.R″ and contains the options for the user to configure the algorithm. Here, it is required to select the disaggregation approach, the number of vehicle categories to which the inventory emissions are attributed, and the choice of either spatial disaggregation or the spatial and temporal distribution of emissions. The second script file is “DROVE.R,” which contains the algorithm and the mathematical functions to perform the disaggregation process; the user does not need to modify the data in this file.

The equation used in the DROVE algorithm by its developers follows those reported by Saide et al. (2009) and adapted by Gómez et al. (2018), and is presented in Equation 1.

The intersection process between the domain grid and the road information allows for an association of the grid cell number for each road segment, ensuring that the same road segment is not accounted for within more than 1 cell of the domain grid.

With this information, in Option 1, the algorithm calculates the total length of road segments for each grid cell in the domain and divides this number by the sum of the length of all road segments, estimating DFs that are then multiplied by the total emission flows. Subsequently, the emission flows for each grid cell are obtained by multiplying the DFs by the total emission of each pollutant, which guarantees mass conservation with respect to the total emission. Finally, as output data, DROVE provides a disaggregation in terms of total emissions per grid cell (in tons/year or kg/day), as well as disaggregated emission results for each vehicle category.

This research proposal addresses the use of DROVE in disaggregation Option 2, where the distribution of the road network is necessary to perform the spatial disaggregation of emissions. The same top-down approach as Option 1 is used but including a weighting factor (fm) for each vehicle category; this factor depends on the road type and attempts to represent differences in traffic flow. The user can define fm values from zero (0) to one (1) for each road type and for each vehicle category, taking care that the sum of the fm values for each vehicle category is 1, to guarantee mass conservation with respect to total emissions. The values of the fm factor can be obtained from the analysis of traffic counts or vehicle flow data. When local traffic data are not available, the default fm values provided by DROVE can be used, which were obtained from the analysis of vehicle flow data reported by Steer Davies Gleave (2017) in the medium-sized city of Manizales, Colombia. Thus far, the presentation of the DROVE tool follows the steps of González et al. (2020); the adjustments made to DROVE as a result of this research work are presented in Section 3.2.3.

Where: DF: disaggregation factor; l: total length of road segments contained in each grid cell j; j: grid cell; n: total number of cells in the domain; k: type of road; i: vehicular category; fm: type of road weight factor; F: traffic flow.

The integrated methodological framework

The proposal links three methodological models with their respective computational models; the workflow sequence, including the introduced interventions, is described in Figure 4 and enables the optimization of emission distribution accuracy across the modeling chain, resulting in a final DROVE simulation represented in GIS that more faithfully reflects urban reality.

Spatially disaggregated emission inventory

Description of traffic flows and estimation of emissions

The starting point is the mobile source emissions inventory corresponding to local private vehicles (Tandil). The fleet consists of a total of 54,305 private vehicles that routinely travel through the local urban area. Based on the counts associated with the IVE methodology, off-peak and daily peak traffic flows were determined using 10 counting sites as a reference. Peak hours are observed at 11:00–12:00, 12:00–13:00 (midday peak) and 16:00–17:00, 17:00–18:00, 18:00–19:00 (afternoon peak); off-peak hours occur at 07:00–08:00, 08:00–09:00, 09:00–10:00 (early morning), 13:00–14:00, 14:00–15:00 (afternoon lull), and 19:00–20:00, 20:00–21:00, 21:00–22:00 (night). The observed values determine road load proportions of 17% for Residential, 32% for Arterial, and 51% for Highway. Figure 5 presents the Arterial, Residential, and Highway road hierarchies, including the counting sites.

The analysis considers two emission time blocks from a typical day. The Morning off-peak period comprises a scenario of commercial, academic, and administrative stability, taken as the baseline or routine of the urban system between 09:00 and 11:00 h; and the Evening Peak, from 17:00 to 19:00 h, represents a scenario of high demand and traffic congestion due to the peak in commercial activities, school dismissals, and the return home. In this way, scenarios of interest for air quality management in intermediate cities with limited geographical extent are evaluated. The results obtained are expressed in concentrations for both time blocks and as daily totals, in kilograms (Table 1).

A 61.2% increase in the distances traveled by vehicles is observed when comparing the Morning off-peak period and the Evening Peak, and the emission of criteria pollutants shows a rise that follows this increase, ranging between 61% and 67%. Since vehicular movement is not a factor isolated from its context but rather the result of urban system activities, this study aims to assign greater value to the daily events that drive changes in traffic flow behavior. Thus, the observed emissions should spatially reflect these contributions.

Road network classification in tandil

As stated in the methodological section, a non-hierarchical k-means technique was applied to the three quantitative variables (commercial density, population density, and educational establishment density). The procedure utilized Euclidean distance, initialization through find seed locations, and no spatial constraints were employed. From this process, three clusters were constructed based on validation criteria (Pseudo F and elbow method). The three resulting groups are visualized in Figure 6B and are characterized by differences across the three dimensions analyzed. Group 1 (n = 4,141) concentrates low-intensity segments: mean population density = 5.0950, mean educational establishment density = 0.3862, and mean commercial density = 0.1392; its low standard deviation and large size indicate that it groups the majority of low-density residential sections. Group 2 (n = 236) gathers the cores of highest intensity and specialization: mean population density = 49.2506, mean educational establishment density = 10.1020, and mean commercial density = 7.4396; identifying corridors and nodes with high commercial and educational concentration. Finally, group 3 (n = 2,007) corresponds to intermediate or mixed areas with mean values: mean population density = 34.9155, mean educational establishment density = 2.4361, and mean commercial density = 1.0814. In explanatory terms, the three variables show relevant R2: population density = 0.70, educational establishment density = 0.63, and commercial density = 0.49, suggesting that the partition captures the population and educational variation more strongly than the commercial one.

It is important to note that all density values are standardized per 100 m of road segment length. For example, an educational establishment density of ≈2.44 indicates that, on average, approximately 2.44 educational facilities are located within the defined buffer area and normalized per 100 m of road segment. This standardization allows for the comparison of segments of different lengths and provides a consistent metric to relate spatial concentration patterns to real urban features such as schools, commercial activities, and population distribution.

Figure 6C represents the intersection between the five road hierarchies (6A) and the three clusters derived from the analyzed variables (6B), resulting in 10 significant classes. Each class combines a land-use typology (cluster 1, 2, or 3) with a functional road category and meets the minimum representativeness condition (≥10 segments). This matrix of combinations allows for the transition from a purely statistical typology to a functional-spatial typology useful for weighting flows and disaggregation factors in the vehicle emissions inventory.

The resulting classification of road network segments is: 11_Ruta corresponds to national or provincial routes; 21_Av1 indicates low-density avenues; 22_Av2 indicates high-density avenues; 23_Av3 corresponds to medium-density avenues; 31_Co l identifies route collectors; 41_Con is a low-density intra-urban connector; 43_Art corresponds to a medium-density intra-urban connector (arterial); 51_Res1 indicates low-density residential; 52_Res2 high-density residential; and 53_Res3 medium-density residential.

The observed spatial patterns are as follows. High-intensity segments: High-intensity segments, corresponding to high-density avenues and high-density residential areas (22_Av2 and 52_Res2), are predominantly located along main avenues and sections of the commercial and financial center. They form linear corridors and nodal clusters and present the highest mean values across the three variables (population density ≈49.25, educational establishment density ≈10.10, and commercial density ≈7.44). These segments represent focal points of population concentration, commercial activity, and educational service. Spatially, they act as structural axes of the urban system, concentrating flows, services, and mobility demand, and therefore higher weighting factors in traffic assignment and emission models.

Low-intensity residential areas (51_Res1 indicates low-density residential, 21_Av1 indicates low-density avenues, and 41_Con is a low-density intra-urban connector): combinations associated with low-intensity residential areas are distributed in a dispersed and fragmented manner, predominating in residential streets and certain sections of avenues that encircle or connect new expansion areas. These classes show low values across the three dimensions (e.g., population density ≈5.10, educational establishment density ≈0.39, and commercial density ≈0.14 for Cluster 1), and their spatial pattern reflects consolidated urban fabrics with low diurnal and nocturnal activity.

Intermediate or mixed areas: Intermediate or mixed segments, including medium-density avenues, medium-density residential areas, and medium-density intra-urban connectors (23_Av3, 53_Res3, and 43_Art), function as transition zones between high-demand corridors and low-intensity neighborhoods. These segments present intermediate means values (population density ≈34.92, educational establishment density ≈2.44, and commercial density ≈1.08), and combine mixed landuses. Their internal heterogeneity makes them key components for local connectivity and short-distance mobility flows.

Routes or collectors: This category, corresponding to national or provincial routes and collector roads (11_Ruta and 31_Col), groups segments with high or medium functional hierarchy but low levels of local urban activity across the analyzed dimensions. Although they play a key role in regional connectivity, their low-density values distinguish them from active urban corridors. In some cases, these segments may correspond to peri-urban or transit-oriented axes (e.g., freight transport routes), where standard allocation approaches may underestimate their contribution to traffic flows and emissions.

Drove model modification. Intervention code

The new code modifies the DROVE algorithm in R within the “DROVE.R″ script. This allows for the inclusion of as many road hierarchies as the user requires from the “UserInterface.R″ file, ensuring the process can be executed while respecting the original methodology of Gonzalez et al. (2020).

In the introduced modifications, the road type variable K (see Equation 1) is redefined in the data loading matrix of the “UserInterface.R″ file. Consequently, the FD factor can adopt the level of disaggregation resulting from the criteria applied in the hierarchical construction of the road network (in this case, as defined in Section 2.3 and Section 3.2.2).

Additionally, a structural modification was incorporated into the code to externalize the weighting by road type into a text file (.txt) named “fm,” which identifies the number and types of roads, as well as the weighting for each road category.

Computational model integration. Comparative analysis: standard vs. adapted disaggregation

To illustrate the main findings of the integrated use of the tools developed, PM 10 emissions will be used given their importance due to health effects in urban spaces. The values found for the criteria pollutants CO, NOx, and SOx are provided in the supplementary documentation so that other users may retrieve information using this integrated methodology for their own analyses.

As a first step, the road categories obtained from the GIS tool analysis (item 3.2.2) are retrieved, and the traffic counts for the flow time blocks are completed within the new road categories. In this stage, traffic count data for National Route 226 and its collectors, provided by the Municipality of Tandil, were used. In this manner, the total emission mass was redistributed according to the following proportions for the 10 Road Types: 11_Ruta: 0.13; 21_Av1: 0.14; 22_Av2: 0.15; 23_Av3: 0.10; 31_Col: 0.07; 41_Con: 0.02; 43_Art: 0.22; 51_Res1: 0.01; 52_Res2: 0.09; 53_Res3: 0.06. Figure 7 displays this distribution.

The emission disaggregation of the vehicle inventory is presented for three spatial resolutions: 500, 250, and 100 m, and for road hierarchies (Arterial, Residential, and Highway) following the standard Option 2 of DROVE, utilizing the hourly vehicle flow allocation fraction for the city of Tandil (Figure 8A). As a contribution to this, results are presented for the modified Option 2 of DROVE to disaggregate into the 10 road hierarchies for the same spatial resolutions (Figure 8B). Concentrations are expressed as normalized values in g/day/km 2.

Based on the descriptive statistical analysis (Table 2) of the emissions presented in Figure 8A compared to 8B, and considering the values per cell for each spatial resolution, a significant increase in data dispersion and maximum values is observed. The standard deviation increases by up to 49% when moving from 3 to 10 categories, this increase in dispersion proves that the model captures the functional heterogeneity of the city, allowing for a distinction between quiet residential streets and dense commercial corridors that were previously treated as equal. The 500 m resolution tends to concentrate mass into fewer cells, raising the absolute maximums per unit. Logically, by grouping more area into a single cell, the mean value per cell unit increases (from 2.05 to 51.33 g/d), and the extreme values for each resolution are accentuated when shifting from 3 to 10 categories. Figure 8A demonstrates that a 500 m resolution generates a false alarm of saturation in isolated cells, whereas the 100 m grid allows for the precise identification of the continuity of critical corridors, which is essential for designing effective environmental zoning plans. Ultimately, the change in disaggregation from 3 to 10 road categories allows for the identification of urban hotspots of interest for mobility management that would otherwise be attenuated. This facilitates progress in the analysis of spatial patterns of urban behavior; an hourly analysis on typical days in small or intermediate cities evaluated with high spatial resolution (500 m or more) would lack meaning without accounting for the heterogeneity of the data—which reflect the urban system—as it would be lost to homogeneity caused by the resolution of the grids used. The spatial disaggregation in Figures 8, 9 will allow for the identification of spatial sectors critical to behavioral patterns and the evaluation of the most immediate actions to be taken.

The differences between both approaches (disaggregation for 3 road types and the proposed model) are also analyzed in relation to spatial behavioral patterns in Tables 3, 4. Section 3.2.2 explained how urban characteristics define and group road hierarchies. The total daily emissions associated with each behavioral pattern (Table 4) allow for the inference of which local road groups should be prioritized for intervention; thus, in this case, the greatest management effort should be directed toward Intermediate/Mixed areas, followed by Routes and collectors’ low density, and High intensity. It is worth noting that 49.7% of emissions occur in only 5.63% of the analyzed segments; therefore, the application of this tool demonstrates that urban interventions should not be sweeping, but rather strategic and localized within these road types to maximize efficiency in resource-constrained contexts.

A spatial reading of the roads associated with the highest emissions will allow for interventions in behavioral patterns, enabling the management of traffic flows either permanently or at specific times associated with peak urban hours. To provide depth to this analysis, the Morning off-peak and Evening peak periods are compared, as represented in Figure 9. Since the previous analysis demonstrated that working with 10 road hierarchies is suitable for addressing urban behavioral patterns related to mobility, the emission disaggregation in this instance is performed directly using this road classification for all three grid resolutions.

The descriptive statistical analysis (Table 5) shows that the temporal increase between Morning off-peak and Evening peak is constant, with values around 17%. This reflects urban activities such as commercial closing times, school departures, or commutes back home. Significant changes are observed in the data dispersion values, which increase as the resolution becomes coarser. In the Evening peak block, the standard deviation (SD) rises from 2.14 at 100 m to 29.28 at 500 m. This suggests that at 500 m, the city’s heterogeneity becomes much more extreme despite using the analysis with 10 road categories; conversely, at 100 m, the mass distribution is more detailed and balanced across the road network. This is also confirmed by the contrast in the maximum values found, where the afternoon peak at the 500 m resolution (429 g) is more than four times higher than at the 100 m resolution (97.54 g). Consistent with the explanation for Figure 8, the 100 m resolution, being finer, not only better isolates the individual contribution of road categories (such as arterials or the route), but also allows us to affirm that there are no substantial changes in the areas affected by traffic during daytime and evening peak hours.

The temporal results highlight two central points. First, they emphasize once again that obtaining total emissions through road hierarchies associated with behavioral patterns is not useful unless accompanied by an adequate resolution—that is, a finer grid consistent with the local spatial scale. Second, when observing traffic behavior in small and intermediate cities, travel distances are shorter; consequently, although there are changes in hourly emission levels, these are not as marked as might be expected (Table 5). This finding challenges the conventional notion of “peak hours” typical of megacities. Traffic management in cities like Tandil must focus more on the spatial location of activities (schools, commercial centers) than on sweeping temporal restrictions, given that the pollutant load remains structural and sustained throughout the day.

The example of Tandil for Figure 9 indicates that connectivity segments (such as collectors or neighborhood entry points) and avenues (associated with schools and commercial centers) act as structural axes of the urban system where emissions do not vanish abruptly but are instead sustained, driven by the activities that characterize the city. Thus, the highest emissions occur in the urban core area associated with commercial activity and the city center, framed by avenues that function as transition bands between high-demand corridors and low-intensity neighborhoods. This spatial interpretation of emissions, which is of interest for public health management, is only possible through the 100 m resolution. The identified structural axes concentrate infrastructure that attracts vulnerable populations, such as schools (children) or healthcare centers (children and the elderly), exposing vulnerable groups to fine particulate matter loads. In this manner, the diagnostic enables the prioritization of mitigation actions within the framework of SDG 11, demonstrating that mobility management in intermediate cities must focus on these structural corridors to effectively reduce environmental impacts and enhance urban resilience.

The main contributions to inventory disaggregation approaches have been developed over the last decade and are not numerous; the objective has been to improve the disaggregation of emissions in spatio-temporal models. Without attempting to be exhaustive, we present a few examples. Jing et al. (2016) simulated the distribution of on-road emissions for off-peak and peak hours in the city of Beijing based on emissions estimated with COPERT at a grid resolution of 1 km × 1 km; spatial emissions were estimated for different roads using traffic flow models fed by vehicle counts and average road speeds. Gómez et al. (2018) performed a spatio-temporal disaggregation for the city of Manizales (350,000 inhabitants) at a resolution of 1 km × 1 km and used simplified traffic count methods that show difficulties by increasing uncertainty in spatial disaggregation; nevertheless, the work disaggregates total vehicle emissions achieved with the IVE model into three road types, which are adjusted with three traffic levels (high, medium, and low) to define traffic flow levels that were successfully used in the disaggregation. Later, González et al. (2020) developed the DROVE computational model with the intention of facilitating the disaggregation of vehicle emissions in contexts of limited information; to this end, they tested their model starting from a base case with robust traffic flow input information and compared the precision of two other disaggregation options with less information (Model Options 1 and 2). The contribution of González et al. (2020) uses medium-sized cities (400,000 inhabitants) and large cities (Bogotá, with more than 8 million inhabitants) as case studies for grid resolutions of 1 km × 1 km and 0.25 km × 0.25 km, with good precision in the results observed when using Option 2.

Viteri et al. (2023) present their work in Ecuador using the disaggregation of an IVE model inventory which, given its spatial scope, simulates emissions on a high-resolution grid (4 km × 4 km) and obtains the emissions using the DROVE computational model by González et al. (2020); these simulations maintain the approach of three strengthened road types while adjusting vehicle flow input data. An approach that incorporates urban dynamics is highlighted by Feng et al. (2023), who emphasize the importance of working with multiple factors to improve emission allocation, including terrain slope, traffic congestion, and population density in their analysis; here, grids with a 1 km × 1 km resolution are utilized. From this overview, we can identify that high spatial resolution is desirable for identifying hotspots, regardless of the size of the city or region. To the best of our knowledge, existing studies do not address small or intermediate cities with smaller spatial and population scales where, as demonstrated here, spatial and temporal disaggregation at high resolutions of 0.25 km × 0.25 km do not faithfully reflect the local distribution of emissions when limited information is available.

In response to this challenge, the present work proposes an integrated methodological framework that optimizes the spatial disaggregation of bottom-up vehicle inventories in contexts of limited information. The strategy combines base estimates obtained through the IVE model with the use of GIS and the adaptation of the open-source Algorithm for Spatial and Temporal Disaggregation of On-road Vehicle Emission Inventories (DROVE). The contribution of this computational tool manages to explain urban behavior in resource-constrained contexts, enabling the disaggregation of emissions by indirectly inferring vehicle activity. By utilizing urban variables, precision in the simulation is gained, which is further enhanced as the spatial resolution of the grid is increased in cities with smaller spatial extensions. As a significant distinction from other studies, it should be noted that this work incorporates socio-demographic factors of interest. These allow for the assignment of a classification to roads based on their urban behavior, categorizing them into road hierarchies dependent on urban patterns for improved simulation in emission disaggregation. Consequently, this enables the identification of urban hotspots of interest for mobility management that would otherwise be attenuated. Thus, for example, a lower data dispersion is observed when using smaller grids (from 500 m to 100 m) even when using the same number of road hierarchies. Thus, with a resolution of 100 m, the emission distribution is more detailed and balanced along the road network, as shown in Figure 8. Precisely, this resolution better isolates the specific contribution of the different road categories.

On the other hand, as observed in Table 5, traffic in intermediate cities like Tandil is characterized by shorter travel distances; consequently, although there are changes in hourly emission levels, these are not as marked as might be expected. In this regard, the conventional notion of “peak hour” typical of megacities must be reconsidered. For traffic management in cities like Tandil, it is crucial to focus more on the spatial location of activities (schools, commercial centers) than on sweeping temporal restrictions, given that the pollutant load remains structural and sustained throughout the day.

This facilitates progress in analyzing spatial patterns of urban behavior by considering the heterogeneity of the data—which reflect the urban system—favoring the adoption of appropriate management measures to protect public health and advance toward SDG 11: Sustainable Cities and Communities.

Among the limitations of this study, it can be noted that the IVE model requires significant time dedication; specifically, it necessitates circuits for surveying driving patterns and at least 1 day of vehicle counting across three road types, which must be reinforced until at least 1 day of counting is reached for each new road class incorporated into the behavioral patterns resulting from the GIS. Furthermore, while typical weekdays are considered here, the study did not explore changes occurring on weekends due to recreational activities or local events associated with a tourist city; the latter could involve emission increases at specific frequencies throughout the year that are worth addressing.

The integration of computational tools and their adaptation to determine spatial emissions in intermediate cities allows for proposals within an environment closer to reality by considering spatial variables and territorial attractors as predictors of vehicle flow on a typical day. Traffic counts were only used to adjust the traffic flow assigned to each road category to modify the hourly flow fractions of DROVE’s Option 2. The DROVE disaggregation algorithm in R was successfully adapted to work with as many road hierarchies as desired to represent the characteristics of the urban system. The incorporation of spatial behavioral patterns, represented by the grouping of road hierarchies, allowed for working with limited economic resources and dispensing with the input of traffic models. However, local knowledge plays a significant role in the analysis to ensure an emission distribution closer to reality. Thus, the use of GIS prior to disaggregation optimizes the application of inventory results for public management, and the adaptation of the DROVE model contributes to representing it accurately in space.

In summary, this work provides an integration of computational models that allows for the development of bottom-up emission inventories with a high-resolution spatial representation. By incorporating and representing the urban system in the analysis, it identifies critical emission hotspots using limited traffic flow data while maintaining an output representative of the local reality.

Understanding the spatial behavior of pollutants enables the identification of usage patterns and activities that require prioritization in transport planning—whether through infrastructure improvements, traffic flow optimization, or the transition toward zero-emission mobility. In this sense, grouping emissions according to road hierarchies allows for a tool integration that offers an innovative approach for microscale inventory assessment.

This information supports decision-making for low-emission policy management because it captures emission profiles associated with driving patterns and speed changes that are relevant for atmospheric pollution mitigation. Consequently, it becomes possible to propose changes in pedestrian zones, establish routes for bike lanes, implement scheduled time windows to pedestrianize streets, or restrict the access of certain vehicle technologies in key urban sectors.

One pollutant studied to demonstrate the utility of the tool is PM 10. Due to its health effects, it is a pollutant that requires continuous monitoring and tracking; high-resolution inventory disaggregations provide information that enables a more precise siting of monitoring equipment for urban air quality networks, the execution of pollutant dispersion modeling with inputs more closely aligned with real-world street conditions, and, as a result, a more reliable approach to models analyzing health effects from exposure to air pollution.

The “smart management” of transport must also encompass small and intermediate cities because they represent a significant share of the global population. Green transport systems for environmental resilience, as a public policy, encounter limitations when small and medium-sized cities lack baseline information to implement management and mitigation measures for urban air pollution. This tool overcomes this lack of information, enabling urban planning policies to prioritize infrastructure, optimize traffic flows, or plan the transition toward zero-emission mobility in a strategic and localized manner.