Formaldehyde column retrieval as a proxy for biogenic VOC emissions
Formaldehyde in the troposphere betrays the oxidation of isoprene and other biogenic VOCs. TROPOMI and GOME-2 retrieve its column daily at continental scale, giving emission modellers a top-down check on bottom-up inventories, provided fire interference is properly removed.
Sensors
- TROPOMI / Sentinel-5P: UV-Vis DOAS spectrometer retrieving HCHO total column at 3.5 × 5.5 km nadir pixel (upgraded from 7 × 3.5 km in August 2019), daily global coverage, spectral fitting window 328–356 nm. Operational L2 HCHO product latency roughly 3 hours after observation.
- GOME-2 / MetOp-A, B and C: Scanning UV-Vis spectrometer, HCHO retrieval at approximately 40 × 80 km pixel (global coverage in 1.5 days per instrument). Three MetOp satellites in polar orbit provide multi-daily sampling; coarser resolution than TROPOMI but an archive extending to 2007.
- OMI / Aura: UV-Vis push-broom spectrometer, HCHO column at roughly 13 × 24 km pixel, daily global coverage. Row-anomaly artefacts reduce effective swath coverage from 2009 onward, but the archive from 2004 underpins long-term trend studies.
- VIIRS / Suomi-NPP and NOAA-20 (ancillary fire data): 375 m active fire detection and fire radiative power (FRP) from the I-band at 3.74 µm, with twice-daily polar overpass. Used as ancillary input to flag and separate pyrogenic HCHO from biogenic HCHO in mixed scenes.
Why formaldehyde is the messenger, not the target
Isoprene (C5H8) is the dominant biogenic VOC emitted by broadleaf forests, grasslands and some crops. Global biogenic isoprene emissions are estimated in the range of 400 to 600 Tg per year, dwarfing anthropogenic VOC sources. Isoprene itself is too reactive and short-lived to accumulate in detectable column amounts from orbit. Formaldehyde, however, is a stable intermediate in its oxidation chain: roughly two to four HCHO molecules are produced per isoprene molecule oxidised, with an atmospheric lifetime of two to four hours under typical daytime photolysis and OH conditions.
That two-to-four-hour lifetime is the key constraint. It is long enough for HCHO to build up to detectable column densities over active vegetation, but short enough that the signal stays geographically close to its source. In practice, the effective spatial footprint of the biogenic HCHO signal is on the order of tens to a few hundred kilometres downwind, depending on boundary-layer wind speed. This makes HCHO a genuinely local proxy, unlike CO or CH4, which can travel hemispheric distances before dissipating.
The UV DOAS retrieval: what the spectrometer actually measures
Differential Optical Absorption Spectroscopy (DOAS) isolates HCHO absorption features in the 328–356 nm window against a reference spectrum, separating the narrow molecular absorption structure from the broad Rayleigh and Mie scattering background. The retrieval yields a slant column density (molecules per cm²), which is then converted to a vertical column density (VCD) using an air mass factor (AMF) computed from a radiative transfer model. AMF uncertainty is the dominant systematic error: it depends on assumed vertical profile shape, surface albedo, and aerosol loading, and can introduce errors of 30 to 50 percent in individual pixels over strongly scattering or aerosol-laden scenes.
TROPOMI's 3.5 × 5.5 km pixel is a genuine advance over its predecessors. At that resolution, individual river valleys with distinct forest composition can be separated, and the signal from isolated deforestation fronts becomes distinguishable from intact canopy. The detection limit for HCHO VCD is approximately 0.5 × 10¹⁵ molecules per cm² for a single overpass, with noise reducing significantly when monthly means are composited. Over the Amazon or Southeast Asian peatlands, peak HCHO columns routinely reach 10 to 20 × 10¹⁵ molecules per cm² during the growing season.
Biomass burning is the interference that cannot be ignored
Pyrogenic emissions are a serious confound. Burning vegetation produces HCHO directly in the fire plume, and also emits other VOCs that oxidise to HCHO downwind. In tropical regions during the dry season, fire-influenced HCHO columns can rival or exceed the biogenic background, and the two sources are spatially collocated: fires burn forests, and forests are the strongest biogenic emitters. A retrieval that does not account for this will overestimate isoprene emission during fire season.
The standard separation approach uses co-located fire radiative power from VIIRS or MODIS to flag pixels with active burning or within a plume corridor defined by back-trajectory analysis. Pixels with FRP above a threshold (commonly 50 MW in published studies) are masked or flagged in the biogenic inversion. Residual pyrogenic contamination in unmasked pixels remains a known source of positive bias, particularly in months like August and September over Amazonia or Borneo. Being honest about this: no current operational product cleanly separates the two signals in real time. The best practice is to treat fire-season retrievals as upper bounds on biogenic flux unless a full chemical transport model inversion is applied.
From column to flux: the top-down inversion
Converting HCHO VCDs to isoprene emission fluxes requires a chemical transport model (CTM) such as GEOS-Chem or MOZART to relate surface emissions to observed column amounts. The yield factor, the number of HCHO molecules produced per isoprene molecule, varies with NOx regime: high-NOx environments suppress HCHO yield, while low-NOx conditions typical of remote forests increase it. Published yield factors range from roughly 1.5 to 3.5 depending on the chemical mechanism used.
Top-down inversions using TROPOMI HCHO have shown discrepancies of 20 to 60 percent between satellite-derived fluxes and the MEGAN (Model of Emissions of Gases and Aerosols from Nature) bottom-up inventory in tropical regions. These discrepancies are not simply measurement error: they reflect genuine uncertainty in leaf-level emission factors, canopy environment parameterisation, and soil moisture stress functions. The satellite data are most valuable precisely because they expose these gaps, giving inventory developers a spatially continuous constraint that ground-based flux tower networks cannot provide at continental scale.
Honest limits: what the data cannot tell you
Cloud cover is the most immediate operational limit. TROPOMI requires a cloud radiance fraction below roughly 0.5 for a usable HCHO retrieval; in persistently cloudy tropical regions such as the Congo Basin or maritime Southeast Asia, monthly cloud-free pixel fractions can drop below 30 percent during the wet season. Compositing over longer periods recovers spatial coverage but loses temporal resolution, which matters for tracking phenological cycles.
Spatial resolution, while much improved with TROPOMI, still limits attribution to specific land-cover types within a mixed pixel. A 3.5 × 5.5 km footprint over a fragmented agricultural-forest mosaic will average across multiple surface types, and the AMF correction assumes a homogeneous scene. Sub-pixel heterogeneity introduces a retrieval bias that is difficult to quantify without high-resolution land-cover ancillary data. Additionally, the HCHO proxy is specific to isoprene and a few monoterpenes; other biogenic VOCs such as methanol or acetaldehyde do not produce HCHO efficiently, so the retrieval captures only part of the total biogenic VOC budget.
Putting it to work in an emission inventory programme
The practical workflow for a government or research client runs as follows. Daily TROPOMI L2 HCHO files are ingested and quality-filtered using the recommended qa_value threshold (typically 0.5, as specified in the ESA product documentation). VIIRS FRP data are collocated to flag fire-influenced observations. Monthly mean VCD maps are computed over the domain of interest, then passed to a CTM inversion to produce gridded isoprene flux estimates at roughly 0.25° resolution. Uncertainty envelopes on the flux are propagated from both retrieval noise and AMF model spread.
Satellize runs this pipeline on open constellations including TROPOMI and VIIRS for clients who need emission inventory constraints without standing up their own satellite data infrastructure. The Tonga crop-estimation programme showed how open-constellation analytics can be operationalised for a small sovereign client; the same data-operations model applies to biogenic VOC monitoring for a forestry ministry or a national greenhouse gas inventory team. The next concrete step for a prospective client is to define the geographic domain and the temporal baseline needed, which determines whether TROPOMI alone suffices or whether the longer GOME-2 and OMI archive is needed to establish a trend.
Typical figures
| Spatial resolution (TROPOMI HCHO) | 3.5 × 5.5 km (nadir, post-August 2019); 7 × 3.5 km before upgrade |
| Spatial resolution (GOME-2) | Approximately 40 × 80 km per pixel; three MetOp instruments available |
| Revisit (TROPOMI) | Daily global coverage; ~14 orbits per day |
| Spectral retrieval window | 328–356 nm (UV DOAS); HCHO absorption features resolved against broadband background |
| Detection limit (HCHO VCD, single overpass) | ~0.5 × 10¹⁵ molecules cm⁻² (TROPOMI); improves with monthly compositing |
| HCHO atmospheric lifetime | 2–4 hours (daytime, photolysis and OH dependent); limits spatial source footprint |
| L2 product latency (TROPOMI) | ~3 hours after observation (near-real-time stream); offline reprocessed product available |
| Archive depth | TROPOMI from November 2017; OMI from 2004; GOME-2/MetOp-A from 2007 |
| Cloud screening threshold | Cloud radiance fraction < 0.5 recommended; tropical wet-season coverage can fall below 30% monthly |
| Ancillary fire data (VIIRS FRP) | 375 m active fire detection, twice-daily overpass; used to mask pyrogenic HCHO contamination |
Analytics Satellize can run
| Monthly mean HCHO VCD maps | Quality-filtered TROPOMI L2 aggregation with cloud and fire masking; AMF-corrected vertical column compositing | GeoTIFF grids at 0.05° or native resolution, monthly cadence, with per-pixel uncertainty layer |
| Fire-contamination flag layer | VIIRS FRP collocation and plume corridor masking using same-day wind-field back-trajectory screening | Binary flag GIS layer co-registered to TROPOMI grid, delivered alongside HCHO composite |
| Isoprene emission flux estimate | Top-down inversion using published HCHO-to-isoprene yield factors and CTM-derived AMF profiles (GEOS-Chem or equivalent published model) | Gridded flux map (molecules cm⁻² s⁻¹) at 0.25° with uncertainty envelope, quarterly report |
| Seasonal emission anomaly detection | Z-score anomaly against multi-year TROPOMI and OMI baseline climatology; phenological cycle detrending | Anomaly alert with magnitude and spatial extent, delivered within 48 hours of monthly composite completion |
| Long-term VOC emission trend (2004–present) | Merged OMI and TROPOMI HCHO time series with inter-sensor bias correction using overlap period (2017–2019) | Annual trend report with per-pixel slope and significance, suitable for national greenhouse gas inventory submission |
| Land-cover stratified emission breakdown | HCHO VCD disaggregation by ESA CCI or Copernicus Global Land Service land-cover class within domain | Tabular emission estimates by vegetation type, with confidence intervals, in CSV and PDF 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.