Secondary pollution mapping via peroxyacetyl nitrate column retrieval
Peroxyacetyl nitrate forms when NOx and VOCs react in sunlight, then drifts far downwind before decomposing into ozone precursors. Thermal infrared sounders can retrieve PAN columns globally, linking distant ozone episodes to specific industrial corridors.
Sensors
- IASI (MetOp-A, -B, -C): Infrared Atmospheric Sounding Interferometer; spectral range 645–2760 cm⁻¹ at 0.5 cm⁻¹ apodised resolution; nadir footprint ~12 km diameter; global coverage twice daily per satellite, giving up to six daily overpasses across the three MetOp platforms combined. PAN is retrieved in the 790–800 cm⁻¹ window. Vertical sensitivity peaks in the mid-troposphere (around 500–700 hPa); near-surface layers are poorly resolved.
- MOPITT (Terra): Measurements Of Pollution In The Troposphere; retrieves CO and, in multispectral mode, provides context for PAN formation chemistry. Nadir footprint 22 km × 22 km; equatorial crossing ~10:30 local time; near-global coverage every 3 days. MOPITT CO profiles help constrain the photochemical age of air masses used in PAN source attribution.
- CrIS (Suomi-NPP, NOAA-20): Cross-track Infrared Sounder; covers 650–2550 cm⁻¹ in three bands at up to 0.625 cm⁻¹ unapodised resolution; footprint ~14 km at nadir; daily global coverage. CrIS PAN retrievals are broadly consistent with IASI but provide an independent cross-check and extend the record as MetOp satellites age. Sensitivity profile similar to IASI: mid-troposphere bias.
- TROPOMI (Sentinel-5P): Does not retrieve PAN directly, but its high-resolution (3.5 × 5.5 km nadir) NO2 and HCHO columns are the primary inputs for diagnosing the NOx/VOC ratios that govern PAN production chemistry. TROPOMI data anchor the source-region characterisation that makes downwind PAN attribution credible.
What PAN actually tells you, and what it does not
Peroxyacetyl nitrate is a secondary pollutant: no stack emits it directly. It forms when acetaldehyde and other carbonyl compounds, produced during VOC oxidation, react with NO2 under ultraviolet irradiation. Its defining property is thermal stability. At temperatures below roughly 15 °C, PAN has an atmospheric lifetime measured in weeks, long enough to cross ocean basins. Warm it above 25 °C and it dissociates in hours, releasing NO2 that feeds ozone production far from any obvious source.
That chemistry makes PAN a tracer of reactive nitrogen export. A column of elevated PAN over the North Pacific, for instance, carries a fingerprint of the Asian industrial corridor that produced it, even though the original NOx emission may be 5,000 kilometres away. The difficulty is that the column measurement integrates the full atmospheric depth. You can see that PAN is present aloft; pinning the vertical layer and the source region requires trajectory modelling and corroborating chemistry, not just the sounder retrieval alone.
How thermal infrared sounders retrieve the PAN column
PAN has a distinctive absorption feature centred near 794 cm⁻¹, within the thermal infrared window that IASI, CrIS and similar sounders cover. Retrieval works by fitting the observed radiance spectrum against a forward radiative-transfer model that includes PAN, ozone, water vapour and other absorbers simultaneously. The result is a partial column or a profile on a coarse pressure grid, with averaging kernels that describe where in the atmosphere the measurement is actually sensitive.
The practical floor matters. IASI PAN retrievals, as documented in peer-reviewed work using the FORLI-PAN and ANNI-PAN algorithms, can detect column amounts of roughly 0.3–0.5 ppbv in the mid-troposphere under favourable conditions: low water vapour, moderate viewing angle, clear sky. Clouds below the PAN layer block the signal entirely. Because PAN absorbs in a spectral region shared with ozone and several other trace gases, the retrieval is underdetermined without tight constraints on temperature profiles, which themselves carry uncertainty from the same sounder. Reported retrieval errors on individual profiles are typically 20–40 percent; monthly regional averages are more reliable.
Attributing downwind ozone to an industrial corridor
The analytical chain from a PAN column to a named emission cluster has several steps, each of which can introduce error. First, elevated PAN columns identified from IASI or CrIS are matched to air-mass back-trajectories, typically computed with HYSPLIT or similar Lagrangian models driven by ERA5 reanalysis winds. Trajectories identify candidate source regions over the preceding 5–15 days, the timescale relevant to PAN transport at cold altitudes.
Second, those candidate regions are cross-referenced against TROPOMI NO2 and formaldehyde columns to confirm that the NOx/VOC mixture capable of producing PAN was actually present. Industrial corridors with persistent NO2 signatures and co-located VOC enhancement are strong candidates. Third, photochemical box models or regional chemistry-transport models estimate how much of the observed downwind PAN is consistent with the candidate source, given the transport time and temperature history. This is where the attribution becomes probabilistic rather than deterministic. A single overpass cannot close the attribution; the case builds over multiple days and overpasses.
For enforcement purposes, the honest statement is this: PAN retrieval from orbit can raise a well-founded hypothesis about which industrial region is exporting reactive nitrogen to a downwind air quality problem. It cannot, on its own, name a specific facility or quantify a stack emission rate. That requires the NO2 and VOC retrievals from higher-resolution instruments, plus ground truth.
The co-absorption problem at coarse footprint
The 12–22 km footprints of current thermal infrared sounders are large enough to contain substantial heterogeneity. Urban heat islands, varying surface emissivity, patchy cloud and sub-footprint temperature gradients all perturb the retrieved radiance. PAN's absorption feature near 794 cm⁻¹ sits adjacent to ozone bands and overlaps with contributions from ethane and other hydrocarbons at concentrations typical of polluted air masses. Retrieval algorithms handle this through simultaneous multi-species fits, but the degeneracy is real: in heavily polluted scenes where you most want to detect PAN, the co-absorber burden is highest.
Water vapour is the dominant interference in the tropics and during summer at mid-latitudes. High humidity reduces sensitivity to PAN to the point where retrievals in tropical boundary layers are essentially uninformative. This is not a solvable problem with current sounder technology; it is a physical constraint. PAN column retrieval is most reliable over dry, cold, high-altitude air masses, which is also where PAN is most stable and transport is most efficient. The technique is well-suited to studying intercontinental pollution export; it is poorly suited to diagnosing local industrial episodes in humid lowland settings.
What a monitoring programme actually looks like
A credible PAN-based attribution service combines three data streams. IASI and CrIS provide the PAN column time series, processed to monthly or seasonal composites to average down retrieval noise. TROPOMI supplies the high-resolution precursor maps that characterise candidate source regions. Trajectory and chemistry modelling ties the two together. The output is not a daily alert; it is a seasonal or annual report identifying which industrial corridors show statistically significant associations with elevated downwind PAN and, by extension, with ozone episodes in receptor regions.
Satellize structures this kind of analysis as a multi-sensor fusion product, drawing on open Copernicus and NASA archives and adding the trajectory and chemistry modelling layer. The approach is analogous to the methodology used in the Tonga crop-estimation programme: open-constellation data processed through a rigorous analytical pipeline to answer a specific policy question. For a government or regulator seeking to attribute a transboundary ozone problem to a foreign industrial source, the deliverable is a documented evidentiary chain, not a single map.
Archive depth is a genuine asset here. IASI MetOp-A data runs from 2007; MOPITT from 2000. Trend analysis over that period can show whether a given industrial corridor's PAN export has grown, stabilised or declined, which is directly relevant to treaty compliance assessment.
Honest limits, summarised
PAN retrieval from thermal infrared sounders is one of the more technically demanding applications in atmospheric remote sensing. The signal is weak, the co-absorber problem is persistent, and the footprint is coarse relative to the emission clusters you want to resolve. Single-profile errors of 20–40 percent are normal. Cloud cover introduces gaps that trajectory analysis must bridge. Near-surface sensitivity is poor, which matters because PAN formed in the boundary layer and lofted to the free troposphere carries a different photochemical history than PAN formed in situ aloft.
None of this makes the technique useless. Used correctly, with appropriate uncertainty quantification and corroborating data, PAN column retrieval provides information that no other current satellite measurement can: a direct, global view of long-range reactive nitrogen transport. For enforcement agencies dealing with transboundary pollution disputes, that view, even with its caveats, is considerably better than no view at all.
Typical figures
| Spatial footprint (IASI) | ~12 km diameter at nadir; 3 × 4 footprint array per scan |
| Spatial footprint (MOPITT) | 22 × 22 km at nadir |
| Spatial footprint (CrIS) | ~14 km diameter at nadir |
| Revisit (IASI, three MetOp satellites combined) | Up to 6 overpasses per day globally; any given location covered 2–4 times daily |
| Spectral window for PAN retrieval | 790–800 cm⁻¹ (12.5–12.7 µm) in thermal infrared |
| Minimum detectable PAN column (mid-troposphere, clear sky) | ~0.3–0.5 ppbv; single-profile retrieval error typically 20–40% |
| Vertical sensitivity peak | 500–700 hPa (mid-troposphere); near-surface layers poorly constrained |
| Cloud sensitivity | Opaque cloud below PAN layer renders retrieval invalid; partial cloud degrades averaging kernels |
| Archive depth | IASI MetOp-A from 2007; MOPITT from 2000; CrIS from 2012 |
| Data latency (near-real-time products) | IASI near-real-time products available within ~3 hours of overpass via EUMETSAT |
Analytics Satellize can run
| Seasonal PAN column composites | Aggregation of IASI or CrIS L2 retrievals to monthly/seasonal grids; outlier-filtered mean with retrieval-error weighting | GeoTIFF or NetCDF grid; accompanying PDF report with uncertainty bounds |
| Industrial corridor PAN export index | Spatial masking of PAN composites over candidate source regions; statistical comparison against background columns using published climatologies | Time-series spreadsheet per corridor; annual trend report |
| Back-trajectory source attribution | HYSPLIT ensemble back-trajectories driven by ERA5; cluster analysis linking elevated downwind PAN to candidate upwind regions | Trajectory density maps (GIS layer); probability table of source-region contributions |
| Precursor fingerprint cross-reference | Co-location of TROPOMI NO2 and HCHO columns with trajectory source regions; NOx/VOC ratio diagnostics following published photochemical regime methods | Multi-panel PDF report; source-region classification (NOx-limited vs VOC-limited) |
| Transboundary reactive nitrogen export assessment | Multi-year PAN trend analysis against IASI/CrIS archive; comparison with reported national emission inventories | Evidentiary dossier suitable for regulatory or treaty-compliance submission |
| Ozone episode precursor linkage | Correlation of downwind ozone episodes (from surface monitoring networks or TROPOMI ozone columns) with antecedent PAN transport events identified from sounder data | Episode-by-episode case summaries; annual statistical summary |
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.