Ground-level ozone precursor attribution to industrial emission clusters
Tropospheric ozone forms from NOx and VOC reactions, not direct emission. Attributing surface ozone episodes to specific industrial clusters requires combining TROPOMI NO2 and HCHO column retrievals with air-mass trajectory modelling, and the method has real resolution limits worth understanding before commissioning work.
Sensors
- Sentinel-5P TROPOMI: Primary workhorse. Delivers NO2 tropospheric columns and HCHO total columns at 3.5 × 5.5 km nadir pixel (reprocessed L2 products from August 2019 onward), daily global coverage. Offline L2 data typically available within 3 hours of overpass. The pixel footprint is the binding constraint for attribution in dense industrial zones where multiple facilities sit within a single grid cell.
- Aura OMI: Provides the long-term (2004 to present) NO2 and HCHO record at roughly 13 × 24 km pixel, valuable for establishing multi-year baseline trends and seasonal ozone-regime climatology before TROPOMI's finer-resolution era. Row-anomaly artefacts affect some swath positions and must be screened.
- GEMS (GEO-KOMPSAT-2B): Geostationary UV-visible spectrometer covering East Asia at roughly 3.5 × 8 km, with hourly daytime revisit. Diurnal NO2 and HCHO sampling resolves the photochemical cycle within a single day, which polar-orbiting sensors cannot. Operational since 2021; data access for non-Korean users is still maturing.
- TEMPO: NASA geostationary instrument covering North America, launched April 2023. Designed for sub-hourly NO2, HCHO and ozone column retrievals at approximately 2 × 4.5 km. Diurnal coverage directly captures the morning NOx build-up that drives afternoon ozone peaks. L2 products are in early validation as of 2024.
Why the precursor ratio matters more than either column alone
Ground-level ozone formation is not proportional to NOx or VOC concentration individually. The chemistry is regime-dependent. In NOx-saturated (VOC-limited) regimes, adding more NOx suppresses ozone; in NOx-limited regimes, it accelerates formation. Identifying which regime governs a given industrial cluster determines whether an emission reduction in NOx, VOCs, or both will actually reduce surface ozone. Getting this wrong produces expensive regulation that achieves nothing measurable.
The formaldehyde-to-NO2 column ratio, commonly written FNR or HCHO/NO2, is the satellite-accessible proxy for ozone-formation regime. Published threshold values from photochemical modelling studies suggest FNR below roughly 1 indicates NOx-saturated conditions, above roughly 2 indicates NOx-limited, with a transitional zone between. These thresholds vary with temperature, season and the specific VOC mix, so they are diagnostic guides rather than hard cutoffs. A refinery cluster with high HCHO and moderate NO2 reads very differently from a steel plant dominated by combustion NOx with low HCHO.
What TROPOMI actually measures, and what it cannot resolve
TROPOMI retrieves NO2 by differential optical absorption spectroscopy in the 405 to 465 nm window. The tropospheric column is separated from the stratospheric contribution using a data-assimilation model, introducing uncertainty that is largest at high latitudes and during dynamically active periods. HCHO columns come from the 328 to 356 nm window. Both retrievals degrade under cloud fractions above roughly 0.3, and the HCHO signal has a lower signal-to-noise ratio than NO2, making it noisier over individual facilities.
The 3.5 × 5.5 km pixel is fine enough to isolate a large, isolated facility but not to separate adjacent plants within a petrochemical complex. A typical refinery, power station and port terminal occupying a 10 km coastal strip will collectively fill two to four TROPOMI pixels with mixed signal. Attribution to individual stacks within that cluster is not achievable from TROPOMI alone. That is not a failure of the method; it is a physical limit that any honest analysis must state upfront.
Trajectory modelling closes the gap between column and surface
Satellite columns are vertically integrated quantities. Surface ozone episodes happen at ground level, often downwind of the precursor source by tens to hundreds of kilometres, hours after emission. Connecting the two requires backward trajectory modelling, typically using HYSPLIT or similar Lagrangian tools driven by reanalysis wind fields, to trace the air mass arriving at a receptor site back to its source region. Combining trajectory frequency analysis with elevated TROPOMI NO2 or HCHO over candidate source clusters provides probabilistic attribution rather than a deterministic source fingerprint.
The approach works best when overpass timing is favourable. TROPOMI crosses the equator around 13:30 local solar time, which is well-timed for capturing the morning NOx build-up before photochemical conversion has dispersed the plume significantly. Geostationary instruments like GEMS and TEMPO improve on this by sampling every hour, allowing the diurnal evolution of the precursor field to be tracked and matched against surface ozone measurements at monitoring stations.
Seasonal and meteorological confounders that attribution must account for
Biogenic VOC emissions, primarily isoprene from vegetation, contribute substantially to HCHO columns in warm months and in regions with significant tree cover. A facility surrounded by forest will show elevated FNR in summer partly because of isoprene, not solely because of industrial VOC release. Separating anthropogenic from biogenic HCHO requires either temporal analysis (comparing weekday versus weekend columns, where industrial activity patterns differ from biogenic ones) or spatial masking against land-cover data.
Boundary-layer height is the other major confounder. On calm, high-pressure days the mixing layer is shallow, concentrating precursors near the surface and producing high column densities even from modest emission rates. On windy days the same emission is diluted into a deeper column. This means column anomalies are not a linear proxy for emission rate without meteorological normalisation. Published methods such as the wind-rotation or mass-balance approaches attempt this correction but require reliable local wind fields, which reanalysis products provide only approximately at the scale of individual facilities.
Combining instruments across time: the archive as a baseline
OMI's record from 2004 gives nearly two decades of NO2 and HCHO data at coarser resolution. Overlapping it with TROPOMI from 2018 onward, after careful inter-calibration, allows analysts to assess whether a cluster's precursor loading has grown, shrunk or shifted spatially. This matters for enforcement contexts where a regulator needs to demonstrate that a facility's contribution has changed relative to a pre-regulation baseline, not just that an episode occurred.
Satellize runs this multi-instrument column analysis as part of its broader emissions-monitoring service, drawing on open Sentinel and NASA archives. The Tonga crop-estimation programme is a different domain, but the underlying workflow of combining multi-source satellite retrievals with atmospheric modelling is the same discipline. For industrial attribution specifically, the output is a probability-weighted source contribution map, not a single confident answer, and any deliverable that claims otherwise is overstating what the physics supports.
What a credible attribution product looks like, and what it does not
A defensible attribution analysis delivers: gridded TROPOMI NO2 and HCHO columns averaged over a defined episode window, cloud-screened and quality-filtered to the recommended qa_value threshold (typically qa > 0.75 for TROPOMI L2); FNR maps with uncertainty bounds; trajectory cluster analysis linking receptor sites to source regions; and a written assessment of which industrial clusters fall within the dominant upwind footprint during the episode period. It does not deliver stack-level emission rates from TROPOMI alone, and it cannot distinguish between two adjacent facilities within a single pixel.
Regulators who need sub-pixel attribution should plan for complementary ground-based or aircraft measurement campaigns. Satellite data is most powerful as a persistent, independent screening tool that flags anomalous periods and candidate source regions, reducing the cost and targeting burden on field campaigns rather than replacing them.
Typical figures
| TROPOMI pixel size (nadir) | 3.5 × 5.5 km (reprocessed products from August 2019; original 2018 products were 3.5 × 7 km) |
| OMI pixel size | 13 × 24 km (nominal nadir, degraded by row anomaly in affected positions) |
| GEMS pixel size | Approximately 3.5 × 8 km over East Asia domain |
| TEMPO pixel size | Approximately 2 × 4.5 km over North American domain |
| Revisit (polar orbiters) | Daily global (TROPOMI, OMI); single overpass per day near 13:30 local solar time |
| Revisit (geostationary) | Hourly daytime (GEMS, TEMPO); enables diurnal photochemical cycle monitoring |
| Cloud screening threshold | Radiance cloud fraction < 0.3 for usable TROPOMI NO2 and HCHO retrievals; higher fractions degrade column accuracy substantially |
| TROPOMI L2 data latency | Offline product typically within 3 hours of overpass; near-real-time product within ~3 hours but lower quality-flag reliability |
| Spectral bands used | NO2: 405–465 nm (DOAS); HCHO: 328–356 nm (DOAS) |
| Archive depth | TROPOMI from May 2018; OMI from October 2004; enables ~20-year combined NO2/HCHO trend analysis with inter-calibration |
Analytics Satellize can run
| Episode NO2 and HCHO column composites | Cloud-screened TROPOMI L2 averaging with qa_value filtering; temporal compositing over defined episode windows | GeoTIFF grids and GIS layer of mean tropospheric NO2 and HCHO columns for the episode period, with pixel-level uncertainty flags |
| FNR (HCHO/NO2) regime classification map | Per-pixel ratio of quality-filtered HCHO and NO2 columns; uncertainty propagation from both retrievals; comparison against published photochemical regime thresholds | Classified raster (NOx-limited / transitional / VOC-limited) with confidence intervals, delivered as GIS layer and PDF summary |
| Backward trajectory source attribution | HYSPLIT or equivalent Lagrangian trajectory modelling driven by ERA5 reanalysis winds; trajectory cluster analysis linked to elevated column anomalies at candidate industrial clusters | Probability-weighted source contribution map per receptor site; tabular summary of dominant upwind cluster contributions during episode hours |
| Multi-year precursor trend assessment | Inter-calibrated OMI and TROPOMI NO2/HCHO time series over defined facility footprints; seasonal decomposition to separate biogenic and anthropogenic HCHO signals | Annual trend report with time-series charts, statistical significance assessment, and comparison against pre/post regulatory baseline periods |
| Weekday/weekend VOC attribution screening | Temporal stratification of TROPOMI HCHO columns by day-of-week to separate anthropogenic industrial VOC from biogenic isoprene signal; paired with land-cover masking | Analytical memo with column anomaly tables and maps distinguishing industrial versus biogenic HCHO contributions by season |
| Diurnal ozone-precursor cycle analysis (GEMS/TEMPO domains) | Hourly GEMS or TEMPO NO2 and HCHO retrievals stacked to build diurnal profiles; correlation with surface ozone monitor data where available | Hourly-resolved precursor time series per cluster; identification of peak build-up windows for targeted inspection or enforcement scheduling |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.