Agricultural ammonia plume mapping and respiratory disease burden

TROPOMI measures atmospheric ammonia columns at 3.5 km resolution, exposing livestock and fertiliser hotspots invisible to ground networks. Secondary PM2.5 chemistry links those columns to respiratory disease burden across downwind populations.

Sensors

Sentinel-5P TROPOMI: NH3 total column at 3.5 × 5.5 km nadir resolution (upgraded from 7 × 3.5 km after August 2019), daily global revisit. Detection limit roughly 1×10¹⁵ molecules/cm²; retrievals degrade significantly under cloud fractions above 0.1 and in humid boundary layers.
IASI (MetOp-A, B and C): Thermal infrared sounder measuring NH3 at ~12 km footprint, twice-daily overpass per satellite (morning and evening). Sensitive to elevated plumes rather than near-surface concentrations; complements TROPOMI by detecting tall industrial and wildfire-related NH3 that TROPOMI can miss.
GEOS-Chem chemical transport model: Not a sensor but the essential adjoint modelling framework used to convert NH3 columns to surface-level concentrations and to attribute secondary PM2.5 (ammonium nitrate, ammonium sulphate) formation to specific source regions. Resolution typically 0.25° × 0.3125° in standard global runs, finer in nested domains.
MODIS / VIIRS aerosol optical depth: Used as a cross-check on secondary aerosol loading downwind of NH3 hotspots. MODIS Terra/Aqua at 10 km (3 km Dark Target), VIIRS at 6 km. Provides the aerosol signal that corroborates model-predicted PM2.5 formation from ammonia precursors.

What a column of ammonia is actually telling you

TROPOMI does not measure surface concentrations. It measures the total vertical column of NH3 between the ground and the top of the atmosphere, expressed in molecules per square centimetre. That distinction matters enormously for health applications. A high column over a feedlot on a calm morning can reflect genuine near-surface accumulation. The same column value over a valley on a windy afternoon might mean the gas has already dispersed to background levels at ground level. The retrieval is also weighted toward the mid-troposphere by the thermal contrast between the surface and the overlying air, so cold, humid days in winter suppress sensitivity precisely when ammonia from housed livestock is still being emitted.

The practical floor is approximately 1×10¹⁵ molecules/cm². Below that, retrieval uncertainty swamps the signal in most humid-climate scenes. Published validation studies comparing TROPOMI against ground-based FTIR and in-situ measurements show biases of 10–30% depending on season and surface type. That is honest enough for source attribution and trend analysis. It is not precise enough to replace a ground monitor for regulatory compliance at a single farm.

From gas to particle: the chemistry that creates the health burden

Ammonia itself is an irritant, but the respiratory disease burden attributed to agricultural NH3 is dominated by what happens after it reacts in the atmosphere. NH3 neutralises sulphuric and nitric acids to form ammonium sulphate and ammonium nitrate particles. These secondary inorganic aerosols are a major component of fine particulate matter (PM2.5) in agricultural regions across Europe, South Asia and the eastern United States. A modelling study using GEOS-Chem found that agricultural NH3 emissions account for a substantial fraction of PM2.5 in the North China Plain during spring, though the precise percentage varies by domain and meteorological year.

The chemistry is not linear. In sulphur-limited regimes, adding more NH3 does not produce proportionally more ammonium sulphate because there is insufficient acid to neutralise. In NOx-rich urban-agricultural fringes, however, the nitrate pathway is highly sensitive to NH3 availability. This means that the same tonne of fertiliser-derived ammonia carries very different PM2.5 consequences depending on where it is emitted. Spatial attribution, not just total-column monitoring, is what turns satellite data into actionable policy.

The spring pulse and why timing matters for exposure assessment

TROPOMI time series over temperate agricultural regions show a consistent seasonal pattern. NH3 columns rise sharply in March and April across northwestern Europe and the US Corn Belt, driven by synthetic fertiliser application to winter wheat and early maize, and by slurry spreading after winter storage. Columns can be three to five times higher than the preceding January baseline in intensive arable zones. A secondary peak sometimes appears in late summer after harvest.

This seasonality has direct implications for population exposure modelling. A person living 20 km downwind of a major pig-farming district receives a disproportionate share of their annual NH3-derived PM2.5 dose in a six-week window. Annual average exposure maps, which regulators typically use, obscure that acute loading. Satellite-derived monthly or weekly column composites, fed into a chemical transport model, can reconstruct the episodic exposure profile that chronic-disease epidemiology increasingly demands.

Translating columns to surface exposure: the GEOS-Chem adjoint approach

The adjoint method runs the chemical transport model backwards from a receptor location, computing the sensitivity of PM2.5 at that point to NH3 emissions at every upwind grid cell. Combining that sensitivity field with TROPOMI-constrained emission estimates yields a source-receptor matrix: which farms, districts or countries are responsible for what fraction of the PM2.5 inhaled by a defined population. This is fundamentally different from forward dispersion modelling, which requires accurate emission inventories as inputs. The adjoint approach uses observed columns to correct those inventories before computing exposure.

Practical limits apply. GEOS-Chem at standard resolution cannot resolve individual farm buildings or even individual villages. Nested domains at 0.5° or finer improve this, but computational cost rises steeply. Cloud gaps in TROPOMI data require either temporal compositing (which smooths out episodic peaks) or gap-filling with IASI retrievals (which have coarser spatial resolution). No current workflow produces a fully gap-free, daily, farm-scale surface concentration field. Honest exposure assessments state their spatial and temporal aggregation clearly.

What governments and health agencies can do with this

The most immediate application is prioritising emission-reduction interventions. Column hotspot maps identify districts where NH3 loading is high and where secondary PM2.5 formation potential is also high, given local NOx and SO2 co-emissions. That intersection is a more defensible basis for targeting low-emission slurry injection subsidies or fertiliser application timing restrictions than national inventory averages alone.

A second application is epidemiological. Satellite-derived exposure estimates can be linked to hospital admission records, primary care data or mortality registries to estimate attributable respiratory burden. This requires careful confounding control, particularly for co-emitted pollutants and socioeconomic factors. The satellite column is an exposure proxy, not a direct health measurement. Satellize's analytics work, including the crop-estimation programme in Tonga, demonstrates how open-constellation data can be processed into policy-grade outputs for governments that lack dense ground-monitoring networks. The same principle applies here: a health ministry with no rural NH3 monitors can still produce a defensible district-level exposure ranking from TROPOMI and GEOS-Chem.

Regulatory applications are growing. The EU National Emission Ceilings Directive requires member states to report NH3 inventories. Satellite columns provide an independent top-down check on those bottom-up estimates, and discrepancies between the two are increasingly cited in Commission compliance reviews.

Honest limits before you commission an analysis

Cloud cover is the primary operational constraint. In tropical and monsoonal climates, monthly cloud-free TROPOMI coverage can fall below 30% of days, making seasonal composites unreliable for the wet season. In those settings, IASI's thermal infrared sensitivity to elevated plumes provides partial coverage but cannot substitute for surface-level exposure estimation.

Terrain complicates everything. Valley topography traps NH3 near the surface and creates column enhancements that look like strong sources but may partly reflect slow vertical mixing. Conversely, coastal and windy upland sites disperse emissions rapidly, suppressing column values even when surface concentrations are locally significant. Any analysis delivered without a meteorological co-variate layer should be treated with caution. Finally, TROPOMI's 3.5 km pixel is large enough to blend a major poultry complex with adjacent arable land. Sub-pixel source attribution requires ancillary land-use data and, in some cases, airborne or high-resolution commercial imagery to confirm what is actually on the ground.

Typical figures

TROPOMI NH3 spatial resolution	3.5 × 5.5 km (nadir, post-August 2019 processor)
TROPOMI revisit	Daily global coverage; single overpass ~13:30 local solar time
IASI footprint	~12 km diameter; two overpasses per satellite per day (MetOp-A, B, C)
Detection floor (TROPOMI NH3)	~1×10¹⁵ molecules/cm²; unreliable below this in humid scenes
Cloud fraction threshold	Retrievals flagged above cloud radiance fraction 0.1; effective clear-sky only
GEOS-Chem standard resolution	0.25° × 0.3125° global; nested domains to ~0.5° or finer
TROPOMI archive depth	May 2018 to present (Sentinel-5P launch November 2017)
IASI archive depth	MetOp-A from 2007; MetOp-B from 2013; MetOp-C from 2019
Latency (TROPOMI near-real-time product)	~3 hours after overpass for NRTI; offline product within 5 days
Spectral bands used	TROPOMI UV-SWIR band 6 (~2305–2385 nm) for NH3; IASI thermal infrared 700–1200 cm⁻¹

Analytics Satellize can run

Monthly NH3 column hotspot map	TROPOMI offline product compositing with cloud-fraction filtering; percentile-based anomaly detection against multi-year baseline	GeoTIFF and GIS layer showing column anomalies by district, with cloud-coverage fraction flagged per pixel
Seasonal emission pulse timeline	Time-series extraction over defined agricultural zones; harmonic regression to isolate spring fertiliser and slurry-spreading signal from interannual variability	CSV and chart report showing weekly mean columns per zone across the agricultural calendar, suitable for ministry briefings
Top-down emission inventory correction	TROPOMI column assimilation into GEOS-Chem to scale bottom-up NH3 emission factors by district; adjoint sensitivity analysis	Corrected district-level NH3 emission estimates with uncertainty bounds, formatted for national inventory reporting
Secondary PM2.5 attribution by source district	GEOS-Chem adjoint source-receptor modelling using TROPOMI-corrected emissions; ammonium nitrate and ammonium sulphate formation pathways	Source-receptor matrix GIS layer: PM2.5 contribution (µg/m³) at receptor populations attributed to upwind NH3 source zones
Population exposure ranking by district	Overlay of modelled surface PM2.5 fields with gridded population data (e.g. WorldPop or national census); seasonal weighting by column-derived emission pulse	Ranked district table of estimated NH3-attributable PM2.5 person-days, with confidence interval reflecting retrieval and model uncertainty
IASI gap-fill composite for cloudy periods	Fusion of IASI NH3 retrievals with TROPOMI clear-sky observations using spatial interpolation constrained by GEOS-Chem transport fields	Monthly blended column raster with provenance flag per pixel (TROPOMI, IASI or model-interpolated)

Sources

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.