Tropospheric NO₂ Patterns in Eastern Canada Using the First Year of TEMPO Observations

Abstract

The Tropospheric Emissions: Monitoring of Pollution (TEMPO) instrument enables an unprecedented assessment of diurnal and community-scale variations in tropospheric nitrogen dioxide (NO₂) across North America. This study presents the first exploratory analysis of NO₂ patterns in eastern Canada, including Ontario, Quebec, and Atlantic provinces, using TEMPO observations. We analyzed TEMPO data gridded at 0.02°×0.02° from September 2023 to August 2024 and compared it with the Tropospheric Monitoring Instrument (TROPOMI) and surface-level measurements from Canadian national regulatory monitors. With the hourly resolution of TEMPO, we observed diurnal trends and hotspots that were not recognized by once-per-day TROPOMI measurements, and pinpointed under-monitored areas. NO₂ in eastern Canada’s eight major metropolitan areas, ports, and industrial cities similarly peaked in early morning and declined in later hours. Still, TEMPO detected variations in their hours of peaks and spikes, seasonal, and weekday-weekend distributions. In Atlantic Canada, correlations between TEMPO and TROPOMI column densities, as well as column-surface alignments, were lower (Spearman’s ρ = 0.41 – 0.53) compared to the Quebec City–Windsor Corridor (Spearman’s ρ = 0.81 – 0.90), primarily due to a wider dynamic range of pollution in the latter region. The two regions’ TEMPO-TROPOMI mean absolute differences were 19.3% and 17.1% respectively. Temporal variations (e.g., a later weekday morning peak in Ontario cities) and TEMPO’s identification of additional under-monitored hotspots provide insights into air quality control planning. Our findings motivate future research using multi-year TEMPO data to investigate atmospheric NO₂ sources, transport, exposure and associated population health impacts in Canada.

Plain Language Summary:

Nitrogen dioxide (NO₂) is an air pollutant primarily associated with fossil fuel combustion that poses a significant health burden. A new satellite instrument, Tropospheric Emissions: Monitoring of Pollution (TEMPO), monitors NO₂ pollution in eastern Canada covering more hours during daytime. Using TEMPO’s first-year observations from September 2023 to August 2024, this study explores NO₂ spatiotemporal patterns in Ontario, Quebec, and the Atlantic provinces. High NO₂ concentrated in large metropolitan areas (e.g., Toronto, Montreal), industrial cities (e.g., Windsor, Sarnia, Saint John), and port communities (e.g., Halifax, Quebec City). TEMPO identified additional municipalities with high NO₂ pollution previously unobserved in suburban and remote areas. We observed annual, seasonal, weekday-weekend, and diurnal NO₂ characteristics, which might be attributed to local traffic, industry, and wind-driven transport. These patterns can provide implications for air pollution control. Our findings also motivate future research on applying TEMPO for studying NO₂ exposure and its impacts in eastern Canada.

1. Introduction

Satellite remote sensing has played a vital role in air pollution monitoring in recent decades (Holloway et al., 2021, 2025). Spaceborne sensors passively detect solar radiances reflected or backscattered through the atmosphere. Based on the characteristics of absorption spectra of gaseous molecules and aerosols, measurements of targeted pollutants are retrieved (Hoff and Christopher, 2009). Through differential optical absorption spectroscopy, a slant column density (SCD) is retrieved, representing the total amount of a gas along the incident light path from the sun to the detector, including scattering and reflection within the atmosphere and at the surface. Vertical column density (VCD) representing ambient air pollution directly above the grid location is obtained from transformation of SCD. This involves computing an air mass factor (AMF) that considers altitude-dependent scattering (Palmer et al., 2001; Nowlan et al., 2025), followed by a stratosphere-troposphere separation process that first estimates a-priori tropospheric enhancements from independent instruments and then applies spatial filtering based on stratospheric and tropospheric AMF ratios (Geddes et al., 2018; González Abad et al., 2024).

Tropospheric Emissions: Monitoring of Pollution (TEMPO) is a satellite instrument launched in April 2023 monitoring a variety of trace gas pollutants over the continental United States (U.S.), northern Mexico, and southern Canada. Regular observations started in August 2023. TEMPO operates with a geostationary orbit that provides approximately hourly available data, compared to the approximately once-per-day revisit cycle of sun-synchronous satellites (Chance et al., 2019; Naeger et al., 2021), thus enabling a more comprehensive understanding of the temporal evolution of atmospheric pollutants during daytime. Comparing hourly column and in-situ surface-level measurements can also inform to what extent TEMPO measurements infer surface air quality distributions. Furthermore, TEMPO’s spatial resolution attains 2 km × 4.5 km at the center of its field of regard (FoR) (33.7°N, 91.7°W) (Naeger et al., 2024), which is finer than its predecessor instruments, such as the Ozone Monitoring Instrument (OMI) and the Tropospheric Monitoring Instrument (TROPOMI). Higher resolution observations enable more detailed assessments of spatial variability at the community level.

During TEMPO’s 20-month baseline operation phase (Phase E) from 2023 October to 2025 June, it measured tropospheric and stratospheric nitrogen dioxide (NO₂), total column ozone (O₃), and total column formaldehyde (HCHO) (Liu et al., 2025; Naeger et al., 2024). In this study, we are interested in tropospheric NO₂ VCD, which is associated predominantly with anthropogenic sources such as traffic and industrial emissions at the ground level, as well as natural sources like soil microbial activity and lightning. NO₂ is one of the criteria air contaminants in Canada (Statistics Canada, 2012), with well-studied acute and chronic impacts on human health (Brook et al., 2007; Manisalidis et al., 2020; Parajuli et al., 2021). Exposure to NO₂ is associated with respiratory and lung diseases, particularly pediatric asthma (Achakulwisut et al. 2019; Anenberg et al. 2022) and chronic obstructive pulmonary disease (Gan et al., 2013), and increases the risks of all-cause hospitalization and mortality (Mills et al., 2015).

TEMPO addresses gaps in surface air pollution monitoring, providing an avenue towards understanding spatiotemporal NO₂ patterns across all communities. Eastern Canada, including Ontario (ON), Quebec (QC), and the Atlantic provinces of Nova Scotia (NS), New Brunswick (NB), Newfoundland and Labrador (NL), and Prince Edward Island (PEI), has sparse surface NO₂ monitoring. NO₂ pollution in eastern Canada is attributed to several factors. First, more than half of the national population resides and around 75% of manufacturing activities in Canada are concentrated in the Quebec City-Windsor (QW) corridor along the southern edges of ON and QC, with major cities including Windsor, London, Hamilton, Kitchener, Toronto, Ottawa, Montreal, and Quebec City (Lévesque, 2010). High commuting demand leads to the densest transportation volume nationally along the corridor. In recent years, population growth has driven urban expansion. Second, the St. Lawrence River connecting the Atlantic and the Great Lakes areas is a busy marine logistics and transportation route (Meng and Comer, 2022). In addition, transboundary air pollution from the northeastern U.S. can affect air quality in eastern Canada given mid-latitude westerlies (Environment and Climate Change Canada (ECCC) and U.S. Environmental Protection Agency (EPA), 2024; Stevens et al, 2024).

We examined the first-year of TEMPO tropospheric NO₂ VCD data covering eastern Canada from 1 September 2023 to 31 August 2024 and specifically addressed the following objectives: a) mapping spatial NO₂ distribution to identify highly-polluted communities; b) visualizing daytime hourly NO₂ trends among the most polluted communities, with seasonal and weekday-weekend variations; c) comparing TEMPO and TROPOMI NO₂ hotspots for surface monitoring gap implications; and d) evaluating TEMPO with TROPOMI and surface-level regulatory monitoring data.

2. Data and Methods

2.1). TEMPO Tropospheric NO₂ products

We accessed TEMPO tropospheric NO₂ VCD data, Version 03 “Provisional”, from the National Aeronautics and Space Administration (NASA)’s “Earthdata Search” platform (NASA, 2024), noting that this version recently completed its provisional validation. At high latitudes, pixel resolution becomes coarser due to a larger view-angle distortion between the Earth’s surface and the satellite’s nadir point. For convenience and consistency in processing, we directly used Level 3 data for analysis. Level 3 is the extension of Level 2, which are geo-referenced, calibrated, quality-assessed, retrieved in standard units, with pixels re-gridded from the native spatial resolution across nine granules to 0.02° × 0.02° (around 2 km × 2 km) (Chance et al., 2013; Zoogman et al., 2017). The re-sampling is based on area-weighted averaging covering all pixels in a complete scan (González Abad et al., 2024; Naeger et al., 2024).

For quality control, we followed the TEMPO Science Team’s recommendations on masking tropospheric NO₂ VCD to minimize uncertainties (González Abad et al., 2024). Using ancillary data, we retained only pixels flagged with good quality (i.e., “main_QA_flag” = 0, effective cloud fraction below 20%, and snow-ice fraction below 5%). These filters account for the removal of outliers, extreme viewing geometry (solar zenith angle (SZA) between incident sunlight and vertical direction above the local grid and viewing zenith angle (VZA) between the local grid position and the satellite’s line of sight), shadowing effects, bright surfaces, and successful AMF calculation from a-priori geophysical information (González Abad et al., 2024). TEMPO tropospheric NO₂ VCD data are expressed in number of molecules per squared centimeters (molec/cm²). Still, good-quality retrievals may contain negative values based on the allowable range of SCD (González Abad et al., 2024).

2.2). Extracting temporal aggregates of TEMPO NO₂

We aggregated images as averages of TEMPO tropospheric NO₂ VCD using different temporal grouping dimensions to assess spatiotemporal variabilities. These attributes included: a) year (from 1 September 2023 to 31 August 2024), b) seasons (fall: September – November 2023; winter: December 2023 – February 2024; spring: March – May 2024; summer: June – August 2024), c) weekdays (Monday – Friday) and weekends (Saturday – Sunday), d) daytime hourly intervals. Besides processing each factor independently, we stratified the results by adding seasons as the secondary grouping dimension when extracting weekday and weekend averages, and similarly, splitting hourly averages specific to seasons and days of week. To reduce striping on these annual and seasonal mean-aggregated images (Nowlan et al., 2025) for better visual interpretation, we have applied a Gaussian filter, which performed horizontal and vertical convolutions sequentially with Gaussian kernels (Pan and Chang, 1992).

Each granule’s scan takes around 6 to 7 minutes (Naeger et al., 2024). Atlantic Canada is situated in the first two (easternmost) granules, while the QW corridor overlaps mainly with the third and fourth granules. We adjusted the binning of daytime hours for all observations accordingly with the “time” variable, which indicates the scanning start time from the first granule (González Abad et al., 2024). Scans starting after the 47-th minute are assigned to the next hour for Atlantic Canada. The same operation applied for the QW corridor with the 34-th minute as the threshold. Also, we accounted for changes in time zones for both regions during the daylight-saving period to adapt to hours in local time.

2.3). Spatiotemporal analysis of TEMPO NO₂ at regional and local scales

We first mapped and examined the annual and seasonal distributions of TEMPO tropospheric NO₂ VCD in the two regional extents (Atlantic Canada: −69.05°W to −52.05°W, 43.05°N to 51.05°N; QW corridor: −86.05°W to −69.05°W, 41.20°N to 48.20°N). Then, we selected municipalities with high observed NO₂ VCD, extracting rectangles bounding their corresponding Canadian census population center boundaries for local analysis. Within each municipality, we mapped NO₂ VCD by over-sampling TEMPO pixels to 0.01° × 0.01° using bilinear interpolation to align with the TROPOMI Level 3 spatial resolution. Spatial interpolation provided an approximation of finer spatial gradients to help represent NO₂ for census units smaller than a TEMPO grid at densely-populated areas, and increase heterogeneity of observations in medium-sized cities.

For intra-municipality analysis, we studied weekday-weekend variations and daytime hourly trends. We calculated pixelwise differences between weekday and weekend mean NO₂ VCDs (Eq. 1), and between consecutive hourly mean NO₂ VCDs (i.e., first-order differences in diurnal time series) (Eq. 2). The weekday-weekend contrast illustrated whether an area had higher pollution during normal working days, and how large this deviation was. The hour-by-hour changes resolved the timing when pollution peaked and increased most rapidly.

Denoting $V$ to represent mean tropospheric NO₂ VCD, $D$ for weekdays, $E$ for weekends, $t$ for any hour, for any TEMPO grid $j$ with center’s coordinates ($x_j,y_j$), and we have:

$${\mathrm{\Delta }\left(x_j,y_i\right)}_{\mathit{weekday-weekend}}={\mathrm{V}}_D\left(x_j,y_j\right)-{\mathrm{V}}_E\left(x_j,y_j\right)$$ Eq. (1)

$${\mathrm{\Delta }\left(x_j,y_j\right)}_{\mathit{hour-by-hour}}={\mathrm{V}}_T\left(x_j,y_j\right)-{\mathrm{V}}_{T-1}\left(x_j,y_j\right)$$ Eq. (2)

Using Canadian census tract boundaries, we obtained area-weighted mean NO₂ and its differences for the weekday-weekend and hourly aggregates. To test whether differences were statistically significant, we applied the Mann-Whitney U Test with a 5% significance level. For areas without census tracts, we used census dissemination areas (DA) instead. DA is the fundamental geographic unit for Canadian census statistics, with around 400 – 700 residents, while a tract (around 5,000 residents) is defined only in cities with an urban core population greater than 50,000 (Statistics Canada, 2023). Furthermore, we obtained a population-weighted average of the tract-level mean NO₂ and its difference for targeted municipalities, based on the 2021 Canadian Census of Population (Statistics Canada, 2023).

2.4). Comparing TEMPO with TROPOMI and surface-level NO₂

To compare TEMPO to precursor satellite measurements, we accessed Level 3 TROPOMI tropospheric NO₂ VCD during the same period using Google Earth Engine (GEE) (European Space Agency, 2024), gridded at 0.01° × 0.01°. We used the offline version that uses more robust inputs of a-priori information for retrievals than the near real-time data. TROPOMI Level 3 applies the same re-gridding logic as TEMPO, proportional to overlapping areas performed using the “bin_spatial” function in HARP (Data harmonization toolset for scientific earth observation data), a command line tool that enables inter-comparison of satellite datasets (S[&]T, The Netherlands, 2024). This TROPOMI dataset on GEE retains pixels with a quality assurance value greater than 0.75, which corresponds to a cloud radiance fraction below 50%, and either a snow cover fraction below 1% or a very small difference between cloud pressure and surface pressure (van Geffen, et al., 2022). For more consistent comparisons, we modified the filtering of TEMPO pixels using “amf_cloud_fraction” that represents the cloud radiance fraction for calculating AMF and adjusted the cut-offs of cloud and snow fractions to 50% and 1% accordingly. We matched and evaluated the TEMPO and TROPOMI images with an absolute scanning time difference within 15 minutes. Moreover, TROPOMI images intersecting or with filtered pixels covering less than 50% of the areas of interest were discarded. For colocation, we re-projected TROPOMI pixels from 0.01° × 0.01° to 0.02° × 0.02° using area-weighted averaging. We refer this dataset as “spatiotemporally colocated” in later sections.

We calculated TROPOMI and TEMPO pixelwise averages respectively from the spatiotemporally colocated images and assessed agreement using Pearson’s ($r$) and Spearman’s ($\rho$) correlation coefficients. Pearson’s r is a commonly used metric, while Spearman’s $\rho$ can characterise correlations for non-normally distributed data and is robust to outliers. Additionally, we reported the mean percentage difference (MPD) between TEMPO and TROPOMI tropospheric NO₂ VCDs per pixel to assess the spatial distribution of discrepancies (Eq. 3). Aggregated observations lower than 1×10¹⁴ molec/cm² (less than 1% of pixels) were excluded to avoid inflating the MPD values.

Denoting $V$ to represent mean tropospheric NO₂ VCD, for any colocated TEMPO and TROPOMI grid $j$ with center’s coordinates ($x_j,y_j$), we define:

$${\mathrm{\Delta }}_j=\left({\mathrm{V}}_{\mathit{TEMPO}}\left(x_j,y_j\right)-{\mathrm{V}}_{\mathit{TROPOMI}}\left(x_j,y_j\right)\right)/{\mathrm{V}}_{\mathit{TROPOMI}}\left(x_j,y_j\right)\times 100\mathrm{\%}$$ Eq. (3)

The agreement between tropospheric column and surface-level NO₂ measurements is of public health and policy interests. In Canada, regulatory monitoring stations of the National Air Pollution Surveillance (NAPS) network measure near real-time air quality and upload data to AirNow (Canadian Council of Ministers of the Environment (CCME), 2019). AirNow was developed by the U.S. EPA as an automated and centralized data management system for U.S. and Canadian ambient pollution monitoring (White et al., 2004). We obtained hourly AirNow data for monitoring stations in the QW corridor (N = 41) and Atlantic Canada (N = 23) from 1 September 2023 to 31 August 2024. Subsequently, we aligned the hourly intervals between TEMPO and AirNow NO₂ measurements, and bilinearly interpolated TEMPO’s observations at the coordinates of these monitoring stations for colocation. Besides correlation coefficients between column and surface NO₂, we plotted their aggregated time series by months, weeks, weekdays, and daytime hours. Since surface measurements are expressed in parts per billion (ppb), we applied min-max normalization to scale TEMPO and AirNow observations respectively to a range from 0 to 1 for comparison.

2.5). Detecting previously unobserved and under-monitored high-NO₂ areas

Since TEMPO covers hours beyond TROPOMI’s once-per-day local overpass time at around 13:30, we hypothesized that TEMPO could capture residential communities of high NO₂ that were unobserved by TROPOMI. We calculated the area-weighted annual mean tropospheric NO₂ VCD (September 2023 – August 2024) from TEMPO and TROPOMI for each census population center in eastern Canada. Then, we ranked the population centers into percentiles by their NO₂ levels. Specifically, we extracted the population centers with TEMPO’s percentile rank exceeding TROPOMI’s percentile rank by 0.50 (i.e., PR_TEMPO (%) – PR_TROPOMI (%) > 50%). To discern if such difference was attributable to pollution observed by TEMPO’s expanded temporal coverage or inconsistencies between TEMPO and TROPOMI, we replicated this analysis using the spatiotemporally colocated dataset.

To identify communities persistently under-monitored at surface level, we examined TEMPO’s and TROPOMI’s annual observations. At the regional scale, we highlighted census population centers within top-decile area-weighted NO₂ VCD. At intra-urban scales, we calculated the bilinearly interpolated satellite measurements at each regulatory monitoring location. If these were smaller than the 98th percentile within the municipality area, the current monitoring network could not effectively cover locations with high extremes where possible surface air quality exceedances may occur. Grids with NO₂ VCD exceeding and monitor(s) nearby below the 98th percentile could be regarded as under-monitored. Again, we repeated the top-decile and 98th percentile mapping using the spatiotemporally colocated dataset, to characterize the spatial variations between TEMPO and TROPOMI derived pollution hotspots when time difference is minimized.

3. Results

3.1. ) Regional NO₂ pollution patterns from TEMPO

Based on TEMPO annual mean (September 2023 – August 2024) tropospheric NO₂ VCD, the pixel-level maximum NO₂ in the QW corridor (8.97×10¹⁵ molec/cm²), occurring in Etobicoke in western Toronto, was more than four times higher compared to that in Atlantic Canada (2.07×10¹⁵ molec/cm²), near the primary highway and railway in east Satin John. Most areas of high NO₂ VCD in the two regions were large urban centers and port communities, while rural northern parts of the domains had low NO₂ (Fig. 1). Regarding seasonal means of NO₂ for the two regions, the fall and winter means were generally higher than those in spring and summer (Fig. S1), primarily due to the NO₂ lifetime being longer during the colder seasons.

Fig. 1.

Spatiotemporal distribution of TEMPO (Tropospheric Emissions: Monitoring of Pollution) tropospheric nitrogen dioxide (NO₂) vertical column density (molec/cm²) (from 2023 September 1 to 2024 August 31) in (a) Atlantic Canada and (b) the Quebec City-Windsor Corridor, aggregated as the annual average from 2023 September 1 to 2024 August 31. Gaussian filter has been applied to the images for strip reduction.

In Atlantic Canada (Fig. 1a), NO₂ hotspots occurred in the three most populated port municipalities (Halifax, Saint John, and Cape Breton). Notably, NO₂ in Cape Breton spread northward to the southwestern corner of Newfoundland. Likewise, we observed a dispersion of pollution extending from Pictou in northern NS, to eastern PEI, and the Magdalen Islands in QC. NO₂ levels in smaller communities at the north shores of Chaleur Bay and near the estuary of the St. Lawrence River, such as Baie-Comeau and Sept-Îles in QC, were higher than in their surrounding areas. During winter, NO₂ spread over the Gulf of St. Lawrence (Fig. S1a). In the QW corridor (Fig. 1b), two major pollution clusters were the Greater Toronto and the Detroit-Windsor metropolitan areas. Along with Port Huron-Sarnia, another trans-border region at the north of Detroit-Windsor, these cities were persistent hotspots in all seasons. Moderately high annual mean NO₂ also occurred in Montreal, as well as Cleveland and Toledo at the south shore of Lake Erie near the U.S.-Canada border. In winter, Quebec City experienced high NO₂, comparable to that in Montreal (Fig. S1b).

3.2). Intra-municipality NO₂ pollution patterns from TEMPO

Considering the spatial distribution of TEMPO annual mean observations (Fig. 1), we selected eight municipalities with high NO₂ levels and a large population for further intra-urban mapping: Halifax and Cape Breton in NS, Saint John in NB, Montreal and Quebec City in QC, Toronto, Windsor, and Sarnia in ON. For larger municipalities, including Halifax, Toronto, Montreal, and Quebec City, high-NO₂ clusters appeared in densely populated urban cores, where census tracts were more compact. High NO₂ in Windsor and Sarnia was concentrated in cross-border areas near the rivers (Fig. 2). Cape Breton comprises several communities, and the highest pollution occurred on the northeastern coast.

Fig. 2.

Annual mean of TEMPO (Tropospheric Emissions: Monitoring of Pollution) tropospheric nitrogen dioxide (NO₂) vertical column density expressed in number of molecules per squared centimeters (molec/cm²) from 2023 September 1 to 2024 August 31 in eight selected Canadian municipalities in Atlantic Canada and the Quebec City-Windsor Corridor with high NO₂ pollution observed from Fig. 1, including (from left to right, top to bottom): (a) Halifax, (b) Saint John, (c) Quebec City, (d) Cape Breton, (e) Montreal, (f) Toronto, (g) Sarnia, (h) Windsor. The polygons indicate the boundary of census tracts, except for census dissemination areas in Cape Breton. The purple-colored circles indicate the locations of surface-level regulatory NO₂ monitoring stations. A spatial smoothing by bilinear interpolation to 0.002° × 0.002° (220 m × 220 m) resolution was applied.

We investigated tract-level NO₂ distributions for each selected municipality stratified by seasons, weekdays and weekends. Based on Table 1, average good-quality observations in winter were much fewer than in other seasons, particularly for Montreal and Quebec City (below 1.5%). Regionally, Newfoundland and areas around the estuary of the St. Lawrence River (Quebec City) had consistently lower fraction of retained pixels compared to surroundings (Fig. S2). Limited observations brought high uncertainties. In other words, results not grouped by seasons are less representative of winter pollution. Still, given available observations, NO₂ levels in Toronto, Windsor, and Quebec City were noticeably higher in winter (Fig. 3). Excluding winter, autumn is the season with the highest NO₂ for Halifax, Montreal, and Sarnia. Moreover, most municipalities maintained relatively low NO₂ in summer, except increases in Cape Breton during weekdays and Sarnia.

Table 1.

Seasonal Distribution (averaged count of pixels retained per each pixel location and its percentage over all hours during that season) of TEMPO (Tropospheric Emissions: Monitoring of Pollution) tropospheric nitrogen dioxide (NO₂) vertical column density (VCD) observations for the eight selected municipalities. Each TEMPO image was screened with an effective cloud fraction below 20% and a snow-ice fraction below 5%. Grouped by municipalities and seasons, the sum of retained observations at each pixel location was calculated, and the number given is the average over all pixels. Correspondingly, the percentage of TEMPO observations was calculated. We defined seasons as i) Fall: September – November 2023; ii) Winter: December 2023 – February 2024; iii) Spring: March – May 2024; iv) Summer: June – August 2024.

	Fall (September - November)				Winter (December - February)				Spring (March - May)				Summer (June - August)
Population Center	Weekday (Mon - Fri)		Weekend (Sat - Sun)		Weekday (Mon - Fri)		Weekend (Sat - Sun)		Weekday (Mon - Fri)		Weekend (Sat - Sun)		Weekday (Mon - Fri)		Weekend (Sat - Sun)
Cape Breton	120.4	17.9%	63.0	23.3%	22.5	3.9%	6.6	2.9%	196.7	24.5%	71.2	22.2%	204.4	27.8%	93.3	31.6%
Halifax	158.4	23.5%	55.9	20.7%	15.9	2.7%	12.9	5.6%	276.5	34.4%	70.3	21.9%	262.3	35.6%	91.6	31.0%
Saint John	157.6	23.4%	59.0	21.8%	46.3	8.0%	12.0	5.2%	277.2	34.5%	68.4	21.3%	275.7	37.5%	104.2	35.3%
Québec City	131.9	15.8%	52.8	15.8%	3.7	0.5%	0.6	0.2%	150.9	14.9%	36.3	9.0%	276.6	29.6%	116.1	31.0%
Montréal	167.9	20.2%	64.0	19.2%	8.1	1.1%	1.4	0.5%	205.5	20.3%	72.4	17.9%	261.7	28.0%	120.7	32.2%
Toronto	201.1	24.1%	85.4	25.7%	53.9	7.4%	21.7	7.5%	293.9	29.0%	69.5	17.2%	279.7	29.9%	124.4	33.2%
Sarnia	166.6	20.0%	75.8	22.8%	82.2	11.4%	13.1	4.5%	349.6	34.5%	92.6	22.9%	307.2	32.8%	144.6	38.5%
Windsor	167.9	20.2%	99.0	29.7%	84.2	11.6%	26.1	9.0%	335.7	33.1%	107.8	26.6%	296.2	31.6%	158.2	42.2%

Fig. 3.

Seasonal variation in TEMPO (Tropospheric Emissions: Monitoring of Pollution) tropospheric nitrogen dioxide (NO₂) vertical column density (VCD) among the eight selected municipalities (from left to right, top to bottom: Halifax, Saint John, Cape Breton, Quebec City, Montreal, Toronto, Sarnia, Windsor) by seasons (columns, from left to right: SON, 2023 September 1 to November 30; DJF, 2023 December 1 to 2024 February 29; MAM, 2024 March 1 to May 31; JJA, 2024 June 1 to August 31). Winter months are excluded due to very limited good-quality observations in some areas. The census tract NO₂ VCDs (1×10¹⁵ molec/cm²) are extracted by area-weighted mean (except using census dissemination areas for Cape Breton). The point represents the median NO₂ levels from all tracts, while each error bar extends to the upper and lower quartiles.

Weekday-weekend differences were statistically significant for most selected municipalities and seasons, except for fall and winter in Saint John, winter and spring in Cape Breton and Quebec City, winter in Halifax, and summer in Sarnia (Table S1). Applying a filter of at least 60 weekday and 24 weekend good-quality observations in each census tract per season, we were unable to compare winter differences for five cities due to insufficient observations (Halifax, Saint John, Quebec City, Montreal, Sarnia). For Toronto and Windsor, though remaining statistically significant after filtering, winter differences became smaller. Furthermore, spatially resolved weekday-weekend MPD indicated most areas, except for rural communities north of Saint John, had higher weekday concentrations (Fig. S3). Saint John and Sarnia had a relatively smaller weekday-weekend difference (up to 20% against 40% – 70% in other municipalities). Weekday and weekend top-decile hotspot locations showed relatively large overlap in Toronto, Montreal, and Windsor; while differed in Quebec City, spreading to the south (western Lévis) during weekdays but northeast during weekends.

Hourly population-weighted averaged NO₂ VCD for all eight selected municipalities are shown on Fig. 4, stratified by weekdays and weekends. Consecutive hourly changes typically indicated when local NO₂ accumulated and depleted (Fig. S4). Weekend NO₂ in all eight municipalities and weekday NO₂ in the Atlantic and QC peaked by 9 a.m. local time. Cities in ON attained weekday maxima 1 hour later at 10 a.m., with a major growth from 8 a.m. to 9 a.m. (11% - 20%) followed by a sustaining increase in the subsequent hour (5% - 6%). After the morning peak, NO₂ VCD gradually decreased and flattened during the afternoon and evening. Still, we observed local characteristics. For example, on weekdays, Sarnia maintained steady NO₂ levels from 9 a.m. until 6 p.m., as did Halifax until 2 p.m. Additionally, Sarnia had a strictly increasing trend after 1 p.m. on weekends. Occasionally, NO₂ levels rebounded in afternoon-to-evening hours, for instance, the weekday NO₂ changes in Halifax and Cape Breton after 6 p.m. Similarly, in the QW corridor, besides the continuous NO₂ rise in Sarnia, more than 10% weekend afternoon NO₂ increases were observed (Toronto: 4 p.m. – 5 p.m.; Windsor: 2 p.m. – 3 p.m. and 5 p.m. – 6 p.m.; Quebec City: 3 p.m. – 4 p.m.). Montreal experienced the greatest rise during weekday afternoons at 5 p.m. – 6 p.m. (+15.4%) among the QW corridor’s municipalities. Further breakdown of diurnal patterns by seasons revealed local pollution characteristics in fall, spring, and summer (Fig. S5 – S7). Other than early morning peaks, drastic spikes were observed on weekday noontime in spring (11 a.m. – 12 p.m.; +64.1%) for Cape Breton and in weekend afternoon (3 p.m. – 4 p.m.) in spring for Saint John (+60.2%) and in fall for Cape Breton (+59.1%).

Fig. 4.

Diurnal hourly variation in TEMPO (Tropospheric Emissions: Monitoring of Pollution) tropospheric nitrogen dioxide (NO₂) vertical column density from 2023 September 1 to 2024 August 31 among the eight selected municipalities: (a) Halifax, (b) Cape Breton, (c) Saint John, (d) Quebec City, (e) Montreal, (f) Toronto, (g) Sarnia, (h) Windsor, stratified by weekdays and weekends. To produce each urban hourly curve, we weighted census tract area-averaged NO₂ vertical column densities by population (except using census dissemination areas for Cape Breton). The marker sizes are proportional to the percentage of observations retained from the quality screening for aggregation, averaged over all pixels in the corresponding municipality and hour.

3.3). Comparison between TEMPO and TROPOMI tropospheric NO₂

With the temporal matching and image filtering criteria, we extracted 39 and 45 TEMPO-TROPOMI paired images for the Atlantic and QW corridor spatial extents accordingly from September 2023 to August 2024. The correlation coefficients between TEMPO and TROPOMI tropospheric NO₂ VCDs in the Atlantic region (ρ = 0.53; r = 0.51) were lower than those in the QW corridor (ρ = 0.90; r = 0.90) (Fig. 5). This is likely driven by the smaller dynamic range of observed values in Atlantic Canada compared to the QW corridor, making random instrument variability more significantly impacting the correlational statistics. Comparing the pixelwise MPD maps (Fig. 5c, 5f) and the averaged MPD over the Annual Leaf Area Index land cover classes derived by the Moderate Resolution Imaging Spectroradiometer (MODIS) (Sulla-Menashe and Friedl, 2018) (Table S2), we found higher retrieved values from TEMPO than TROPOMI in urban and built-up areas (Atlantic Canada: +18.8%; QW corridor: +25.8%). Additionally, TEMPO observed higher NO₂ in the sparsely populated rural areas in central-northern ON and QC, and around the Gulf of St. Lawrence. However, over water bodies, TEMPO detected lower NO₂ levels compared to TROPOMI (Atlantic Canada: −11.1%; QW corridor: −12.1%). Averaged over pixels where TEMPO detected higher NO₂ VCD, the MPDs were 19.3% for Atlantic Canada and 17.9% for the QW corridor. Conversely, on pixels with smaller TEMPO VCD, the magnitudes were 19.4% for Atlantic Canada and 16.5% for the QW corridor. Overall, the absolute MPDs were 19.3% (Atlantic Canada) and 17.1% (QW corridor) respectively. Moreover, we investigated the relationship between TEMPO-TROPOMI residuals and land cover classes using linear regression adjusted with terrain height and surface albedo (Table S3). It reaffirmed that TEMPO residuals were positively biased compared to TROPOMI on urban land cover but negatively biased compared to TROPOMI over water, with statistical significance. The magnitude of these biases were larger in the QW corridor. Vegetation and terrain height only had small impacts on residuals. Interestingly, surface albedo and TEMPO-TROPOMI residuals was positively associated in Atlantic Canada (β = 0.89, CI: [0.87, 0.91]), but negatively associated in the QW corridor (β = −0.49, CI: [−0.52, −0.45]).

Fig. 5.

Comparison between TEMPO (Tropospheric Emissions: Monitoring of Pollution) to TROPOMI (Tropospheric Monitoring Instrument) tropospheric nitrogen dioxide (NO₂) vertical column densities (VCD), based on the paired images with scanning time differences within 15 minutes. The graph separates the outputs for (a) – (c) Atlantic Canada and (d) – (f) Quebec City-Windsor Corridor. The left column shows the regression plots with Pearson’s and Spearman’s correlation coefficients, intercepts, and slopes between the TEMPO and TROPOMI NO₂ VCD, the middle column indicates the residual plots (actual TROPOMI VCD against the input values of TEMPO VCD) of the regression line, while the right column illustrates the spatial percentage differences: ∆ = (TEMPO NO₂ – TROPOMI NO₂) / TROPOMI NO₂ * 100%, per each pixel.

3.4). Comparison between TEMPO and surface-level NO₂

The two regions of interest showed dissimilar trends in TEMPO tropospheric NO₂ and AirNow surface-level measurements by monthly, weekly, day-of-week, and daytime hourly averages (Fig. 6). The monitoring data in the QW corridor attained high correlations with TEMPO observations (ρ = 0.81 – 0.88; r = 0.75 – 0.80), while in Atlantic Canada these were moderate (ρ = 0.41 – 0.45; r = 0.35 – 0.52). Moreover, the monthly and weekly averages of AirNow and TEMPO observations in the QW corridor were more closely aligned with each other (Table S4). Observations for the first few weeks during January and February should be interpreted with caution, due to few good-quality observations from TEMPO. For the QW corridor, column and surface NO₂ showed concurrent increases on Tuesday and Friday, with minima on weekends. Sub-daily changes in surface NO₂ preceded TEMPO column NO₂ by approximately two hours. For example, the peak in surface NO₂ at 12 p.m. UTC may correspond to that in TEMPO VCD at 2 p.m. UTC. Both measurements decreased during later hours. Surface NO₂ flattened at around 3 p.m. – 4 p.m. UTC, while column NO₂ flattened at 5 p.m. UTC. Residual patterns generally did not exhibit any trend but appeared slightly more dispersed at higher TEMPO VCD ranges for the daily and hourly comparison in the QW corridor (Fig. S8).

Fig. 6.

Comparison between spatiotemporally-colocated TEMPO (Tropospheric Emissions: Monitoring of Pollution) tropospheric nitrogen dioxide (NO₂) vertical column density and surface NO₂ measurements from regulatory monitoring stations (data obtained through AirNow) in (a) – (d) Atlantic Canada and (e) – (f) Quebec City-Windsor Corridor. The four columns show comparisons at (from left to right): monthly mean (starting from September in 2023 to August in 2024); weekly mean (starting from the 36^th week in 2023 to the 35^th week in 2024); day-of-week mean (0 = Monday; 6 = Sunday); and diurnal hourly mean. For each column, the first row shows the regression plots with Spearman’s correlation coefficients between TEMPO and surface NO₂, while due to their differences in physical units, the second row shows the normalized values of the two data sources varied by time.

3.5). TEMPO-derived under-monitored NO₂ hotspots

Eight population centers plus one First Nations reserve in Atlantic Canada had percentile ranks (PRs) of TEMPO’s annual average greater than that of TROPOMI by a magnitude of 0.50 (i.e., 50%) (Table S5; Fig. 7). These communities were mainly located at the north shores near the estuary of the St. Lawrence River and the Chaleur Bay in QC, with Channel-Port aux Basques at the southwestern corner of Newfoundland. For their NO₂ VCDs, seven out of nine were situated in TEMPO’s upper-third (above the 67-th percentile), while all fell within TROPOMI’s bottom 30%. Fig. S9 and S10 replicate the regional and municipality maps in Fig. 1 and 2 using TROPOMI annual mean tropospheric NO₂ VCD. Moreover, the nearest ground-based monitoring stations were 50 km – 300 km away from these population centers. TEMPO-TROPOMI PR differences from the spatiotemporally colocated dataset of these coastal hotspots were narrower than those from the annual averages, except for the First Nations reserve (Table S5). The consistent positive bias of TEMPO compared to TROPOMI in built-up areas (as identified in section 3.3) might account for communities already having large spatiotemporally colocated PR differences (>40%). This might be attributed to instrumental specifications (e.g., orbits, viewing geometry, spectral coverage, etc.) and processing algorithms between TEMPO and TROPOMI. In contrast, we found communities with small TEMPO-TROPOMI PR differences or much lower TEMPO PRs in the spatiotemporally colocated dataset (Port-Cartier, Carleton-sur-mer, Maria, Channel-Port aux Basques). Instrumental biases were unlikely to fully explain their large annual PR variations, implying that TEMPO’s expanded daytime monitoring could help capture previously unobserved pollution. Understanding the causes of these differences is beyond the scope of our analysis but we encourage future research on this topic, especially in communities where TEMPO-TROPOMI PR differences are large.

Fig. 7.

Population centers in the Atlantic Canada region with the difference in percentile ranks of tropospheric nitrogen dioxide (NO₂) vertical column density from TEMPO (Tropospheric Emissions: Monitoring of Pollution) and TROPOMI (Tropospheric Monitoring Instrument) greater than 0.5, i.e., PR_TEMPO – PR_TROPOMI > 0.5, highlighted by red-colored polygons and annotated with labels in the map.

Top-decile population centers derived from TEMPO’s first year NO₂ VCD agreed with TROPOMI observations in major populated and industrial cities, while showing dissimilarities in suburban communities (Fig. 8). Under-monitored communities are those with high NO₂ and without any monitor in the neighbourhood. For Cape Breton and Saint John (Fig. 8a), monitors were primarily located in central areas (i.e., Sydney in Cape Breton and the port of Saint John). Three persistently unmonitored communities around Sydney were located on the northern coast, which experienced higher TEMPO NO₂ levels (Fig. 2). The northeastern suburbs of Saint John also had high NO₂, possibly impacted by emission sources from urban Saint John, but lack surface monitoring to characterize the pollution levels. TEMPO additionally identified Channel-Port aux Basques, NL and Cap-aux-Meules, QC as under-monitored communities. In the QW corridor, a cluster of remote communities without nearby monitors in the south of Windsor had consistently high NO₂ VCD (Fig. 8b). Suburbs located in the northeast of Montreal were recently under-monitored from 2023 – 2024 TEMPO and TROPOMI averages. Meanwhile, TEMPO uniquely identified pollution in Woodstock, a mid-sized town between London and Kitchener in Ontario, but did not capture TROPOMI’s monitoring gaps in southern Niagara and Montreal. Within the eight selected municipalities, TEMPO and TROPOMI showed the largest 98th percentile overlap in western Toronto and near Windsor (downtown Detroit) (Fig. 9). In northeastern Cape Breton, TEMPO detected higher NO₂ over lands, against TROPOMI’s NO₂ spread over marine areas. Similarly, the two instruments were less consistent around urban cores in Montreal, Quebec City, Saint John, and the peninsula of Halifax. In Sarnia, TROPOMI detected hotspots closer to northern major residential zones, while TEMPO indicated top extreme in the southern areas.

Fig. 8.

Population centers in the (a) Atlantic Canada region and (b) the Quebec City-Windsor Corridor identified as hotspots (exceeding the 90th percentile). The colour classifications were defined based on the population center’s area-weighted tropospheric nitrogen dioxide (NO₂) vertical column density (VCD) annual average from TEMPO (Tropospheric Emissions: Monitoring of Pollution) and TROPOMI (Tropospheric Monitoring Instrument) (from 2023 September 1 to 2024 August 31). Brown color indicates common hotspots from TEMPO and TROPOMI measurements during 2023 – 2024. Red and blue colors refer to those identified only from TEMPO and TROPOMI measurements. Additionally, the grey-colored points represent the current NO₂ regulatory monitoring locations.

Fig. 9.

Intra-urban hotspots (exceeding the 98th percentile) in the eight selected Canadian municipalities in Atlantic Canada and the Quebec City-Windsor Corridor with high NO₂ pollution observed from Fig. 1, including (from left to right, top to bottom): (a) Halifax, (b) Saint John, (c) Quebec City, (d) Cape Breton, (e) Montreal, (f) Toronto, (g) Sarnia, (h) Windsor. The colour classifications were defined based on the annual mean tropospheric nitrogen dioxide (NO₂) vertical column density (VCD) from TEMPO (Tropospheric Emissions: Monitoring of Pollution) and TROPOMI (Tropospheric Monitoring Instrument) from 2023 September 1 to 2024 August 31. Light-brown and slate-blue coloured areas were derived from annual TEMPO and TROPOMI measurements respectively. Points indicate current NO₂ regulatory monitoring locations. Correspondingly, light-green, dark-red and cyan coloured points indicate whether the station is placed within the TEMPO and TROPOMI, TEMPO-specific, and TROPOMI-specific hotspots. The spatial resolution for intra-municipality mapping is 0.01° × 0.01°.

Compared to the annual average dataset, hotspots mapped with the spatiotemporally colocated dataset exhibited similar patterns regionally (agreement in most highly-polluted cities with suburban variations), but less overlap for intra-urban settings except the Detroit-Windsor area (Fig. S11 and S12). Variations between TEMPO and TROPOMI hotspots in Fig. 8 and Fig. 9 dataset might be resulted from a mix of factors, such as biases between TEMPO and TROPOMI retrievals varied by land characteristics (section 3.3), difference in daytime coverages of the two instruments, and other factors not yet understood. To further clarify intra-urban NO₂ differences and strengthen the spatiotemporally colocated analysis, more ground-based measurements are needed.

4. Discussion

TEMPO advances high-resolution NO₂ monitoring, facilitating research on atmospheric processes, emissions, and health risks. In this study, we conducted an exploratory analysis on spatiotemporal patterns of first-year TEMPO tropospheric NO₂, and validated its agreement with TROPOMI and surface monitor data in eastern Canada. Specifically, we calculated population-weighted NO₂ concentrations, accounting for diurnal, seasonal, weekday-weekend, and community-level spatial variability. Compared to TROPOMI, which is limited to a once-per-day overpass at approximately 13:30 local time, TEMPO observed NO₂ during morning rush hours when traffic-related air pollution is more severe. TEMPO might also be more sensitive to local emissions in built-up areas. New TEMPO observations helped identify previously unrecognized high-pollution areas in eastern Canada that are currently not covered by the surface air pollution monitoring network. Comparing TEMPO NO₂ VCDs in monitored and unmonitored areas can support optimizing monitor site selection and better evaluate the monitor network’s confidence to capture extreme pollution scenarios. Typically, this type of analysis relies on model-derived surface concentrations (Wang et al., 2024).

The eight municipalities with high TEMPO NO₂ levels were either densely populated urban centers or industrial hubs. Toronto and Montreal are the two largest metropolitan areas, each occupied by more than four million residents. Halifax is the most populous city in Atlantic Canada. Large cities have high traffic volumes and power consumption contributing to NO₂ emissions. Furthermore, most selected municipalities are close to marine transport routes. Saint John, Windsor and Sarnia have heavy industry, such as chemical and oil refineries, pulp and paper mills, automotives, etc., with the latter two situated on the U.S. border and impacted by transborder point and mobile source emissions.

Annually predominant southwesterly winds might be associated with observations among the three Atlantic port communities (Fig. S13). Northeastern rural areas of Halifax (i.e., Lake Major) (Fig. 2) and suburbs of Saint John (i.e., Quispamsis-Rothesay and Hampton) (Fig. 8) were identified with high NO₂ from TEMPO. Despite being less populated, Cape Breton possesses three coal-fired or petroleum-based power plants. The prevailing winds might carry NO₂ emitted northeast leading to the emergence of TEMPO hotspots in New Waterford and Glace Bay, instead of its more populated center in Sydney, NS (Fig. 2 and S10), which illustrated a typical mismatch between observed hotspot and monitored locations (Fig. 9d). Furthermore, northeastern transport of NO₂ emissions from Cape Breton across the Cabot Strait might impact southwestern Newfoundland (Channel-Port aux Basques) (Table S5). Interestingly, a year-round ferry service operates between Cape Breton and Channel-Port aux Basques. The ferry terminal is located on the northwestern coast of Cape Breton (Sydney Mines), where TEMPO observed high NO₂ but not TROPOMI. The scheduled routine departure at 11:45 a.m. and summer-specific departure at 5:30 p.m. also coincided with the annual noontime and summer weekday evening NO₂ rises in Cape Breton (Fig. 4 and S5). Moderately high NO₂ levels were also observed along ferry routes connecting Pictou in NS, eastern PEI, and Magdalen Islands in QC. Still, more evidence is required to confirm whether these were coincidences and quantify impacts from wind-driven transport and shipping-related NO₂ emissions in Atlantic Canada.

Among communities with much higher TEMPO PRs (Fig. 7; Table S5), Baie-Comeau and Sept-Îles are most populous stopover ports for transoceanic goods-carrying ships, with aluminum processing industry (Adebayo et al., 2014; Ferrario et al., 2022). With iron ore mining activities, Sept-Îles is also a connection point for railway and truck transport. For communities by the Chaleur Bay, pollution sources are mainly found on the south shore in NB, including petroleum-fuelled power and chemical processing plants (Fraser et al., 2011), which might contribute to cross-boundary pollution in QC. To our knowledge, public health outcomes associated with NO₂ exposure in these regions have not been well explored.

Differences between TEMPO and TROPOMI observations might also offer new insights into intra-urban NO₂ pollution sources. TEMPO’s 98th percentile specifically covered the North End neighbourhood of Halifax, where major railroads, a shipyard, and cross-harbour bridges exist (Fig. 9a). The harbourside monitor was placed outside of both the highest polluted areas identified by TEMPO and TROPOMI. In Quebec City, three monitoring stations are located south of the periphery of TEMPO’s 98-percentile areas, unable to capture NO₂ extremes from high densities of railroads, highways, and an industrial park within that area (Fig. 9c). For Montreal, TEMPO’s high-NO₂ area was more north compared to those based on TROPOMI (Fig. 2; S10), coinciding with industrial zones near northern Longueuil and Montreal-Est (Fig. 9e). In Sarnia, TEMPO detected the highest NO₂ in the southwest, which is the city’s manufacturing zone, known as the “Chemical Valley”, accommodating 40% of Canada’s chemical production (Atari et al., 2008) (Fig. 2 and 9g). Local monitoring expansion is recommended around southern Sarnia to characterize exposure. In Toronto, the highest NO₂ occurred near the international airport and proximal warehousing and distribution facilities (Fig. 9f). Areas surrounding warehouses in the U.S. had 20% higher NO₂ VCD based on TROPOMI (Kerr et al., 2024), while similar investigations are lacking in Canada. Woodstock, a top-decile hotspot identified specifically by TEMPO, is a manufacturing center for automotive assembling. In particular, Saint John and Montreal have regulatory monitors situated within TROPOMI’s and TEMPO’s 98th percentile areas respectively (Fig. 9b; 9e). Examining the AirNow surface NO₂ data at these monitors during the study period, we found that for Saint John, the monitor within TEMPO’s 98th percentile recorded both higher daytime (7 a.m. – 6 p.m.) and early-afternoon overpass (1 p.m. – 2 p.m.) hourly averages (3.75 ppb and 3.35 ppb) than the monitor within TROPOMI’s 98th percentile (3.35 ppb and 3.29 ppb). However, for Montreal, the daytime average was lower at the monitor within TEMPO’s intra-urban hotspots (7.11 ppb). Proximity to major highways (Autoroute 40 and Route 138) could lead to higher measurements by the other two surface monitors captured by TROPOMI (8.81 ppb and 11.78 ppb).

The diurnal trend observed using TEMPO matches the general trends of previous findings from in situ ground monitors. In Halifax, nitrogen oxides (NOx) emissions based on measurements in 2017 also peaked in the early morning and then declined gradually (Mitchell et al., 2021). Ozone photochemistry might be an important factor accounting for seasonal NO₂ variations. TEMPO observed lower weekday NO₂ levels in Halifax and Cape Breton in spring than in summer (Fig. S5). Prior research found a spring peak in ground-level ozone in NS (Mitchell et al., 2021), during which more NO₂ would be dissociated by sunlight. In contrast, ground-level ozone in the QW corridor peaked in summer (Brook et al., 2014), when TEMPO observed minimum NO₂, except for Sarnia (Fig. 3). Persistent release of pollutants from chemical production might account for Sarnia’s distinct summer NO₂ diurnal patterns (Fig. S5). Sarnia’s afternoon increasing trend on weekend might align with the longer cross-border waiting time at the Blue Water Bridge (the U.S.-Canada crossing connected to Sarnia) during weekends after 2 p.m. (Maoh et al., 2018), which leads to elevated emissions from trucks. Moreover, the weekday morning peak in Toronto during summer appeared late (11 a.m.). Studies suggested in summer, photolysis causes a drop in the early morning NO₂, and conversion of nitrogen monoxide (NO) into NO₂ by ozone inflates satellite-observed NO₂ VCD in the late morning (Edwards et al., 2024).

TEMPO-TROPOMI and column-surface correlations in Atlantic Canada were significantly lower than in the QW corridor (Fig. 5 and 6). Background NO₂ in Atlantic Canada is lower than in the QW corridor, due to the lack of highly polluted large metropolitan areas like Toronto, Detroit–Windsor, and Montreal. Higher sensitivity and precision are required to discern smaller ranges in NO₂ levels. Situated at the eastern-edge granules and high latitudes, VCD in Atlantic Canada is subject to additional atmospheric corrections due to large SZA and VZA (Vanhellemont et al., 2014). Moreover, coarser spatial resolution at the edges of the FoR may lead to lower agreements with other observations due to over-smoothing. Analyzing SCD might be another research direction leading to other insights of NO₂ distributions. Moreover, TEMPO treated a pixel as either land or water, and used the same wind-dependent climatological parameters to model water surface albedo (Nowlan et al., 2025), while TROPOMI considered mixed land cover pixels and sea ice conditions when adjusting surface albedo (van Geffen, et al., 2022). These differences might be associated with the greater contrast between urban and water retrievals from TEMPO compared to TROPOMI (Table S2), and opposite relationships between surface albedo and TEMPO-TROPOMI residuals in the two regions (Table S3). Further research into where and why these residuals happen is needed.

This study has a few limitations that warrant future investigation. First, for Canada’s high latitudes and cold weather, applying low cutoffs of cloud and snow fractions could mask a large proportion of pixels during winter months (Table 1). However, winter NO₂ pollution is usually more severe because of the reduced daylight suppressing NO₂ photolysis, shallower atmospheric boundary layer mixing, and increased fossil fuel combustion for residential or commercial heating (Brook et al., 2014; Mitchell et al., 2021). Recent research also found higher surface NO₂ during cloudy-day than clear-sky conditions (Goldberg et al., 2025). Imbalance in percentages of good-quality observations across seasons could bias spatial patterns when comparing annual aggregates regionally and locally, particularly less representative of municipalities with fewer good-quality winter retrievals, such as Quebec City and Montreal. Seasonal expansion of surface monitoring could fill data gaps and balance the estimation of surface exposure. Similarly, sub-daily uncertainties are higher at the sunrise and sunset hours (e.g., 7 a.m. and 6 p.m.) because AMF calculations depend on MODIS observations (overpass at 10:30 a.m. and 1:30 p.m.) (González Abad et al., 2024). High temporal-resolution surface monitoring could be targeted during these hours. Second, from our results, current TEMPO NO₂ retrievals have a larger contrast between built-up land and water compared to TROPOMI. Future studies may consider adjustments in processing algorithms or calibrating performance for offshore and rural NO₂ detection. Third, we only analyzed TEMPO’s first-year data, which might be prone to single-year anomalies. While our analysis is exploratory in nature, our findings can still statistically inform general spatiotemporal patterns, potential uncertainties, and how TEMPO-derived hotspots may differ from past understanding. As TEMPO has planned for 15-year operational lifetime (Zoogman et al., 2017), long-term analyses spanning multiple years of TEMPO measurements are necessary to assess diurnal variability and persistence of hotspots. This will also help thoroughly evaluate TEMPO’s various biases against other satellite and ground-based instruments that strengthen the evidence for monitoring policy and public health applications. Lastly, for surface measurements, AirNow only stores real-time data from NAPS monitors without validation. Validated data would be released one to two years later, and should be incorporated in future comparisons. Alternatively, the Pandora global network offers ground-based observations, though limited to Toronto and Windsor (United Nations Environment Program Ozone Secretariat, 2024). Pandora gives columnar density with similar retrieval algorithms as satellite-based measurements, while NAPS monitors provide surface measurements based on chemiluminescence techniques and cavity attenuated phase shift spectroscopy, which might overestimate concentrations due to interference from reactive nitrates (Dunlea et al., 2007).

5. Conclusion

Our work summarized spatiotemporal patterns of tropospheric NO₂ in eastern Canada. Leveraging TEMPO’s high-resolution data enabled us to examine diurnal variations, pinpoint population centers with previously unobserved NO₂ pollution, and assess intra-urban hotspots not covered by the regulatory monitoring network. The findings on diurnal variations could potentially shape a more adaptable NO₂ control policy. For instance, by examining regions experiencing traffic congestion when peaks and rapid increases in NO₂ occurred, authorities might design or revamp transit and road systems considering emission mitigation. Our study makes an important first step towards better understanding NO₂ pollution and its association with emission sources, atmospheric transport, and health outcomes in eastern Canada. Next-step studies may focus on uncertainty of TEMPO observations, such as surface pollution in winter and variations by land characteristics.

Key Points:

• TEMPO observed diurnal NO₂ variations in eight major polluted areas in eastern Canada, generally with a morning peak, mid-day decline and flattening.

• TEMPO identified previously unrecognized NO₂ hotspots and under-monitored suburban communities.

• Good-quality observations were scarce in winter, bringing high uncertainty.

• TEMPO-TROPOMI and column-surface correlations were weaker in Atlantic Canada than in southern Ontario and Quebec, due to a smaller dynamic range of the observations.

Data Availability Statement:

All data used in in this manuscript’s analyses are publicly available. TEMPO Level 3 tropospheric NO₂ VCD and its support data (provisional V03) were accessed from https://www.earthdata.nasa.gov/ and maintained by NASA’s Atmospheric Science Data Center (DOI: https://doi.org/10.5067/IS-40e/TEMPO/NO2_L3.003) (Gorelick et al., 2017; NASA, 2024). TROPOMI Level 3 tropospheric NO₂ VCD was downloaded from the Google Earth Engine (GEE), collection “COPERNICUS/S5P/OFFL/L3_NO2”, which is owned by the European Space Agency (European Space Agency, 2024). Similarly, the MODIS Type 3 (Annual Leaf Area Index) land use classification data were downloaded from GEE, collection “MODIS/061/MCD12Q1” (Sulla-Menashe and Friedl, 2018). Near real-time Canadian surface monitoring data were downloaded from AirNow through “PyRSIG” (GPL-3.0 license specified in the repository https://github.com/barronh/pyrsig), a Python package for the Remote Sensing Information Gateway maintained by U.S. EPA (https://epa.gov/rsig) (Henderson, 2025; U.S. EPA, 2024). All codes for the data extraction and visualization are implemented on Python (version >= 3.11) and made available in the repository: https://github.com/tksiu/tempo-no2-east-CA with MIT license.