Background & Summary

Integrating diverse datasets spanning daily weather, traffic, and air pollution across six major Norwegian cities presents a significant endeavor to understand the complex interactions within urban environments1. Urban areas face increasing challenges related to climate change, air quality decline, and traffic congestion, necessitating comprehensive data analysis for informed decision-making and policy formulation2.

Intense traffic congestion exacerbates air quality issues by increasing traffic volume3,4. Air quality hazards pose significant challenges for urban residents and city administrations alike5,6. Research has shown that air pollution, including nitrogen oxide (NOx) and particulate matter, significantly contributes to mortality rates among urban populations7,8. These emissions primarily originate from transportation sources9, making the transportation sector crucial in mitigating air pollution10. Additionally, meteorological factors such as temperature, precipitation, wind speed, relative humidity, and atmospheric pressure significantly influence the dispersion and accumulation of particulate matter less than 2.5 micrometers in diameter(PM2.5)6. Therefore, research on the interplay between air pollution and meteorological factors affecting chronic respiratory diseases presents particular challenges in Northern Europe. As one of the Nordic countries, Norway is narrow and vast due to its geographical characteristics. Each city is located in different geographical locations, thus traffic flow and weather conditions vary significantly11,12. Tromsø in the north is situated within the Arctic Circle, experiencing a cold climate with reduced sunshine. Fredrikstad in the southeast is influenced by the North Sea climate, receiving more rainfall throughout the year and being prone to storms. Bergen in the west maintains high humidity year-round and is frequently battered by Atlantic storms. The capital, Oslo, located in the south, enjoys four distinct seasons and a relatively mild climate. Trondheim in the center is inland and has a colder climate. Stavanger in the southwest is influenced by the North Atlantic Warm Current, resulting in relatively warm winters but precipitation throughout the year. The differing traffic and weather conditions in these cities illustrate Norway’s diverse geography and its profound impact on residents’ lives.

This study compiles data from the Norwegian Public Roads Administration, the Norwegian Institute of Air Research, and the Norwegian Meteorological Institute, supplying a comprehensive dataset covering a decade of observations from 2009 to 2018. The dataset encompasses various pollutants such as Nitric oxide (NO), Nitrogen dioxide (NO2), NOx, PM2.5, and Particulate Matter less than 10 micrometers in diameter (PM10), along with detailed weather and traffic information. Through meticulous site selection and verification, a total of 18 monitoring stations were identified across the cities of Oslo, Bergen, Trondheim, Fredrikstad, Stavanger, and Tromsø.

The broader goal motivating the creation of this dataset is to facilitate multidisciplinary research endeavors and inform evidence-based policymaking. By synthesizing data from multiple sources, this study aims to clarify the complex interplay between weather patterns, traffic dynamics, and air quality in urban settings. The potential reuse value of this dataset is substantial., offering researchers and policymakers a comprehensive resource to investigate environmental trends, assess the efficacy of mitigation strategies, and develop sustainable urban planning initiatives.

This dataset holds promise for a wide range of applications, including epidemiological studies linking environmental exposures to health outcomes, transportation planning to alleviate traffic congestion, and climate modeling to assess the impact of urbanization on local weather patterns13. Since this dataset consists of daily data, previous studies have shown that exposure to high concentrations of NOx and PM2.5, increases the incidence of cardiovascular and respiratory diseases, with a lagged effect based on the day14. Similar studies include Chossière et al.15, who found that high NOx concentrations contribute to mortality15, and Feng et al.16, who pointed out that even lower concentrations of PM2.5 pose a risk to public health16. Therefore, the dataset is of great value in exploring the health consequences caused by traffic pollution. Overall, the integration of these datasets represents a crucial step towards enhancing our understanding of urban environmental dynamics and advancing strategies for fostering healthier, more sustainable cities17.

Methods

The methodology involves several steps to generate and compile the dataset, including experimental design, data collection, analysis, and computational processing. Traffic flow, air pollution, and weather data are sourced from different providers. Compiling the dataset involved meticulous verification of geographical locations to identify common monitoring stations across the three types of data. Initially, the ten largest cities in Norway were selected based on population. Weather, pollution, and traffic data were collected from local monitoring stations for each city, prioritizing traffic and air pollution monitoring stations.

The geographical coordinates of traffic and air pollution monitoring sites were cross-validated to identify the closest matching sites, ensuring accuracy. This process was iteratively repeated for each city, resulting in the selection of monitoring stations such as Kirkeveien, Smestad, E6 Manglerud, and Alnabru in Oslo.

Weather variable selection involved excluding cities lacking certain weather conditions. The locations of weather monitoring stations were compared with air pollution and traffic monitoring stations to identify the nearest ones, iteratively repeating this process. Due to the absence of identical monitoring stations, especially since the distribution of weather monitoring stations differs significantly from those of the other two public data management agencies, I opted to select geographically proximate locations between weather and the other two. I assumed that the detailed weather conditions observed at these stations would be similar. Given Norway’s relatively small urban area, the meteorological conditions in areas close to traffic and pollution are also expected to be similar. Future studies could consider employing spatial network analysis for a more precise exploration, as highlighted by previous research18,19. For instance, Kang et al.20 utilized spatial analysis and geoAI in their study conducted in Sweden20. This approach enhances the capability to examine spatial relationships and geographic influences effectively. The paper provides specific distances to clarify the degree of proximity between these locations in Table 1 and Fig. 1. Figure 1 shows the locations of monitoring stations, and red hollow circles indicate the monitoring stations. Additionally, Fig. 2 outlines the steps in data collection and processing, offering a clearer and more intuitive understanding.

Table 1 List of stations and their distances.
Full size table
Fig. 1
figure 1

Map showing the locations of monitoring stations, sourced from a Google map, Map data ©2024 Google.

Full size image
Fig. 2
figure 2

The steps for data collection and processing.

Full size image

The selected analysis period covers daily data from 2009 to 2018. The dataset includes maximum and minimum normalized data after data imputation using the R package “mice”, ensuring no missing values. Data from Oslo, Bergen, and Stavanger spanning 2007 to 2018 were collected using the same methodology for subsequent analysis.

Given that data collection occurred just as COVID-19 was concluding or nearing its conclusion, it’s crucial to acknowledge that traffic patterns and associated air pollution likely fluctuated during the pandemic. Therefore, data from the period before COVID-19 was selected for analysis. Additionally, annual air pollution data undergoes quality assurance procedures before July 1 of the following year; before this date, adjustments may occur due to ongoing quality control measures. Additionally, hourly data from Oslo’s Bygdøy for the first half of 2019 was collected, with missing data on May 17th supplemented.

To avoid overcrowding the horizontal axis with daily data points, the data was initially aggregated into monthly intervals. Subsequently, selected data points were sampled for visualization. Figure 3 presents visualizations for traffic volume (TV) and air pollution data in Oslo, as well as NOx values from several other chosen cities.

Fig. 3
figure 3

Visualization of aggregated monthly data for Oslo’s traffic and air pollution, along with NOx values from selected cities.

Full size image

Data Records

The data presented in the repository consists of comprehensive information on daily weather, traffic, and air pollution across six major Norwegian cities over ten years, from 2009 to 2018. The dataset was compiled from multiple sources, including the Norwegian Public Roads Administration, the Norwegian Institute of Air Research, and the Norwegian Meteorological Institute.

The dataset is available at the Scientific Data Bank platform at the following link: https://www.scidb.cn/en/detail?dataSetId=5eb39369148d4f40bb5ec398d6503c4621. The data is provided in Excel or CSV format, ensuring compatibility with various analysis tools and software packages. Here’s a summary of the data records:

  1. 1.

    Data Content: The dataset includes information on common pollutants such as NOx, NO, PM2.5, etc, along with detailed weather variables including air temperature, dew point temperature, cloud cover, wind speed, humidity, precipitation, and snow cover.

  2. 2.

    Repository Structure: The repository contains separate files for each city, with data organized in a tabular format. Each file corresponds to a specific monitoring station within the respective city.

  3. 3.

    File Naming Convention: The files are named according to the city and the type of data source. For example, “Oslo_traffic.csv” contains traffic data from Oslo.

  4. 4.

    Data Availability: The dataset covers daily observations for each variable from 2009 to 2018.

  5. 5.

    Data Validation: The dataset underwent thorough validation to ensure accuracy and consistency. This involved meticulous site selection, verification of geographical locations, and comparison of monitoring stations across different data sources. Missing values were imputed using the R package “mice”.

  6. 6.

    Code Availability: The code used for data analysis and processing is available on the author’s GitHub repository at the following link: https://github.com/congca/Scientific-data-publishand https://doi.org/10.5281/zenodo.11649590. Readers can navigate the repository to access specific city data files, explore the variables of interest, and utilize the dataset for various research purposes, including environmental studies, urban planning, and policy formulation.

Tables 2 to 3, and Appendix 1 provide comprehensive descriptions of the variables (columns) included in the traffic, air pollution, and weather datasets. These tables offer detailed information about each variable, including their names, units of measurement, and the specific data they represent. In Fig. 4 heatmap, the color blue denotes the presence of variables, whereas white indicates their absence. This visualization offers a clear representation of the occurrence of different weather variables across each city. They clarify the abbreviations used in the datasets and enumerate the specific weather-related variables available for each city. It is important to note that all cities include data variables related to both traffic and air pollution, ensuring a consistent dataset for comparative analysis and research across different urban areas. Table 4 presents a detailed breakdown of urban and countryside locations within various cities. This table categorizes specific monitoring sites and their corresponding environmental contexts, providing valuable insights into the spatial distribution of data collection points. By distinguishing between urban and countryside locations, this table facilitates a deeper understanding of how geographic context influences traffic and air pollution data. The variable abbreviations are provided in Table 5.

Table 2 Variable descriptions for “Traffic.csv” files.
Full size table
Table 3 Description of variables in the “pollution.csv” file.
Full size table
Fig. 4
figure 4

Weather variables in different cities.

Full size image
Table 4 The urban and countryside locations in various cities.
Full size table
Table 5 Abbreviations and Full Terms.
Full size table

Technical Validation

Compiling the final dataset proved time-consuming due to the disparate sources of traffic, pollution, and weather data. The objective was to identify common monitoring stations or, at least, those near each other across the three types of data sources. As a result, the process involved meticulous verification and comparison of geographical locations using Google Maps.

Initially, data collection began by selecting the ten largest cities in Norway based on population. Subsequently, data from local weather, pollution, and traffic monitoring stations were gathered from the respective government websites for each city. Each city boasted numerous monitoring stations for each type of data. The process prioritized traffic and air pollution monitoring stations.

To ensure accuracy, the geographical coordinates of traffic and air pollution monitoring sites were cross-referenced and verified against each other, thereby identifying the closest matching sites. This iterative process was repeated for every city, culminating in the selection of monitoring sites such as Kirkeveien, Smestad, E6 Manglerud, and Alnabru in Oslo; Danmarks Plass, Loddefjord, Rådal, and Åsane in Bergen; Kannik and Schancheholen in Stavanger; and Bakke Kirke, E6 Tiller ved XXL, and Elgeseter in Trondheim.

Additionally, the air pollutant monitoring stations adjacent to the highway were chosen to exclusively study traffic-related air pollution, mitigating other potential sources of pollution. This meticulous site selection process resulted in a total of 18 monitoring stations across various cities, with the earliest available data dating back to February 1, 1996, provided in hourly intervals. The common pollutants monitored across these sites include NO, NO2, NOx, PM2.5, and PM10, underscoring the significance of the selected monitoring sites and the effort invested in their identification. The information provided here serves as a testament to the value derived from the time-intensive process of identifying these monitoring sites, updated as of July 20, 2022.

Following that, the selection process turned to weather variables. However, the weather data collection website operates with a system wherein variables are chosen before selecting the city, resulting in a reversed order of selection. Given that there are more weather variables than pollutants and traffic, if all the stations in a particular city lack certain weather conditions, that city is excluded. Consequently, only six cities remained in the dataset. Stavanger was initially included but was subsequently removed due to a significant proportion of missing values in certain weather variables upon download.

The subsequent step involved comparing the locations of weather monitoring stations with air pollution and traffic monitoring stations to identify the nearest one. This process was repeated iteratively. The next consideration was to check the time to find the highest frequency but longest period available. The air quality data availability by city and pollutant is in Fig. 5.

Fig. 5
figure 5

Air Quality Data Availability by City and Pollutant.

Full size image

The period chosen is from 2009 to 2018, with daily data. All data is stored in Excel or CSV format and can be found at Scientific data bank. Tables in the data records section provide essential details about the main data inputs and variables. The data underwent maximum and minimum normalization, followed by data imputation using the R package “mice” to ensure no missing values. Additionally, data was collected from Oslo, Bergen, and Stavanger spanning 2007 to 2018 using the same methodology for further analysis. Monthly plot covering traffic volume (TV), NOx, and PM2.5 readings from Oslo, Bergen, Trondheim, and Tromsø over ten years is also available.

Two standard check

Studies have utilized this dataset for related research. For instance, Cao (2024) examined hourly data from the same source to investigate interactions among air pollution, traffic, and weather, comparing machine learning algorithms with traditional methods1. Furthermore, Cao et al. (2024) utilized a subset of this dataset to analyze key meteorological and traffic factors influencing air pollutants, assessing the predictive accuracy of physics-based deep learning and long short-term memory (LSTM)17.

All the data sources are freely available. Air pollution data is obtained from https://luftkvalitet.nilu.no/en/historical. The website offers historical air quality data verified by municipal authorities and the National Air Quality Reference Laboratory, including preliminary data for the current year. You can view charts or download data in CSV format, select specific dates, times, stations, or components, and choose from various averages. Data can be accessed at http://api.nilu.no for personal use and should be properly referenced in your work. When referencing NILU air quality data for Oslo in 2023, cite it as NILU - Norwegian Institute for Air Research. (2023). Air Quality Data for Oslo, January 2023 - December 2023. Available at: http://api.nilu.no.

Weather data is sourced from: https://seklima.met.no/observations/. When using meteorological data provided by the Norwegian Climate Service Center (https://klimaservicesenter.no/), please ensure that you accurately cite them as the source in your research or course materials. This includes listing the Norwegian Climate Service Center in your references and providing a link to their website, allowing readers or viewers to verify the data source.

Traffic flow data is obtained from the Norwegian Public Roads Administration’s website, specifically from https://trafikkdata.atlas.vegvesen.no/#/kart?lat=62.59346885386541&lon=10.86809745774079&trafficType=vehicle&zoom=4.

Use of the traffic data API is governed by the Norwegian Open Government Data License (NLOD), allowing copying, distribution, modification, and commercial use, with considerations for data quality, potential errors, and omissions. The Norwegian Public Roads Administration does not guarantee data accuracy or relevance and assumes no liability for any losses from its use.