Seasonal methane flux mapping over rice paddy agriculture
Flooded rice paddies produce methane through anaerobic decomposition, with emissions peaking mid-season. Combining Sentinel-1 inundation maps with TROPOMI column retrievals lets analysts track the seasonal flux cycle, though separating paddy signal from co-located wetlands and livestock remains genuinely hard.
Sensors
- Sentinel-5P TROPOMI: XCH4 column retrievals at 5.5 × 7 km pixel footprint (upgraded from original 7 × 3.5 km in 2019), daily global coverage, detection sensitivity around 1–2 ppb against a ~1,900 ppb background; cloud fraction threshold typically <0.3 limits usable observations during monsoon cloud cover.
- Sentinel-1 SAR (C-band, 5.405 GHz): Interferometric Wide Swath mode delivers 10 m resolution, 250 km swath, 6-day revisit at the equator with both satellites. Backscatter drop over standing water reliably maps inundated paddy extent through cloud and at night; distinguishes flooded from non-flooded fields but cannot directly measure water depth.
- MODIS Terra/Aqua: 250 m NDVI and 500 m land-surface products at 1–2 day revisit provide phenological timing of transplanting, heading and harvest across large regions. MODIS burned-area and thermal products also flag field burning events that confound methane attribution.
- GOSAT-2: Column-averaged XCH4 at roughly 9.7 km diameter footprint with ~3 day sub-satellite repeat; higher precision than TROPOMI for individual soundings but sparse spatial sampling. Useful for cross-validation of TROPOMI retrievals over the same regional domain.
Why rice paddies are a methane source worth mapping from orbit
Flooded rice paddies create anaerobic soil conditions in which methanogenic archaea decompose organic matter and release methane. The gas escapes partly through plant aerenchyma tissue, partly by ebullition, and partly by diffusion through the water surface. Emission rates vary by water management regime, soil carbon content, cultivar and temperature, making paddy agriculture one of the more variable diffuse sources in the agricultural methane budget.
The seasonal structure is what makes satellite observation tractable. Fields flood at transplanting, remain inundated through most of the growing season, and drain before harvest. That cycle is visible in SAR backscatter and produces a predictable methane enhancement signal in atmospheric columns above major paddy regions. The enhancement is not enormous: regional column anomalies of a few parts per billion above background are typical, which sits right at the edge of TROPOMI's detection capability for individual overpasses. Aggregating many clear-sky observations across a season is what makes the signal statistically separable.
What a flooded field looks like to a radar
Sentinel-1 C-band SAR is the practical workhorse for paddy inundation mapping. Smooth open water returns very little backscatter to the sensor, so flooded fields appear as dark patches in VV or VH polarisation imagery. The contrast against dry or vegetated land is sharp enough that automated thresholding or supervised classification can map inundated extent at 10 m resolution across an entire river delta in a single acquisition.
The complication is that rice canopy grows rapidly after transplanting. Within three to four weeks, emerging vegetation begins to increase backscatter, partially masking the water signal. Time-series analysis across multiple acquisitions, rather than single-date classification, is therefore necessary to capture the full inundation calendar. Published work on the Mekong and Irrawaddy deltas has demonstrated that Sentinel-1 six-day revisit is sufficient to resolve transplanting and drainage dates to within one acquisition interval, which is adequate for seasonal flux modelling.
SAR cannot measure water depth, distinguish irrigation water from rainfall-driven flooding, or identify the crop variety being grown. Those gaps matter for flux estimation because water management practices such as mid-season drainage, which some farmers use to reduce methane emissions, are not directly visible in backscatter alone.
Reading the methane column above a mosaic of sources
TROPOMI measures the total atmospheric column of methane expressed as a dry-air mole fraction. A single 5.5 × 7 km pixel integrates signal from the surface up through the troposphere, which means that anything emitting methane within that column contributes to the retrieval. In lowland agricultural landscapes, paddies, natural wetlands, livestock, and biomass burning can all occupy the same footprint simultaneously.
Source separation is the central analytical challenge. The standard approach is atmospheric inversion: a transport model carries prior emission estimates forward in time and space, then adjusts those estimates to minimise the mismatch between modelled and observed columns. The quality of the result depends heavily on the quality of the prior spatial emission map, the density of clear-sky TROPOMI observations, and the accuracy of the meteorological wind fields used by the transport model. Published regional inversions over South and Southeast Asia, including work using earlier GOSAT data and more recent TROPOMI records, have shown that paddy methane can be statistically attributed when the inundation extent is well constrained by SAR. The uncertainty on seasonal totals remains large, typically tens of percent, and is dominated by source ambiguity rather than retrieval noise.
Cloud cover is a serious practical limit during the Asian monsoon. TROPOMI requires cloud fractions below roughly 0.3 for a usable XCH4 retrieval. In the Indo-Gangetic Plain or the Mekong Delta during the main wet-season crop, cloud-free overpasses can be sparse for weeks at a time. Analysts must either accept gaps in the time series or use gap-filling methods that introduce their own assumptions.
Flux inversion: from column anomaly to emission estimate
Converting a column enhancement into a surface flux requires an atmospheric transport model. Published studies have used both global models such as GEOS-Chem and regional frameworks at finer resolution to invert TROPOMI and GOSAT observations over Asian paddy regions. The general finding is that satellite-derived inversions produce seasonal paddy flux estimates that are broadly consistent with bottom-up inventories such as those underpinning EDGAR, but with substantial regional disagreements that the satellite data help to identify.
A practical workflow for a national programme runs roughly as follows. Sentinel-1 time series establishes inundated paddy area and the transplanting-to-drainage calendar at field scale. MODIS phenology confirms crop stage timing. TROPOMI column anomalies, filtered for cloud and quality flags, are aggregated into weekly or monthly regional means. A transport model, initialised with the SAR-derived emission map as a spatial prior, is then run in inversion mode to adjust flux magnitudes until modelled columns match observations. The output is a seasonally resolved emission estimate with associated uncertainty bounds, expressed in tonnes of CH4 per hectare per season or as a national total.
The honest caveat is that this workflow produces estimates, not measurements. The inversion is underdetermined: many combinations of surface fluxes can produce the same observed column. Additional constraints from ground-based flux chambers, eddy covariance towers, or aircraft campaigns substantially reduce that ambiguity, but most governments do not have dense in-country measurement networks.
Where this fits in a national methane reporting framework
Rice paddy methane is a Tier 1 or Tier 2 source in IPCC greenhouse gas inventory methodology, meaning that most countries currently estimate it from activity data and emission factors rather than atmospheric observation. Satellite-derived flux estimates offer a way to cross-check those inventory figures independently, which is increasingly relevant as countries submit updated Nationally Determined Contributions under the Paris Agreement.
The practical value is not replacing the inventory but auditing it. A TROPOMI-based seasonal analysis can flag whether a country's reported paddy area and water management practices are consistent with the observed atmospheric signal. It cannot resolve individual farms or even individual districts at TROPOMI's 5.5 km footprint, but it can identify regional discrepancies large enough to warrant investigation. Satellize runs this type of multi-sensor seasonal analysis on open constellations, combining Sentinel-1 inundation mapping with TROPOMI column processing, as part of its emissions-monitoring service portfolio. The Tonga crop-estimation programme demonstrated the same underlying approach of using SAR extent mapping to anchor an agricultural signal that would otherwise be uninterpretable from column data alone.
Governments considering satellite-based methane reporting should be clear about what they are buying. A seasonal paddy flux product is a statistical estimate with meaningful uncertainty. It is most defensible when combined with a ground-truth campaign, even a modest one, and when the methodology is documented transparently enough to survive independent review.
Typical figures
| TROPOMI XCH4 pixel footprint | 5.5 × 7 km (post-August 2019 upgrade from 7 × 3.5 km) |
| TROPOMI revisit | Daily global coverage; usable clear-sky observations limited by cloud fraction threshold <0.3 |
| TROPOMI XCH4 precision (single sounding) | ~1–2 ppb against ~1,900 ppb background; regional anomalies detectable by aggregation |
| Sentinel-1 SAR inundation mapping resolution | 10 m (Interferometric Wide Swath mode, VV/VH polarisation) |
| Sentinel-1 revisit at equator | 6 days (both satellites combined); 12 days per satellite |
| MODIS phenology product resolution | 250 m (NDVI), 500 m (land cover), 1–2 day revisit |
| GOSAT-2 footprint | ~9.7 km diameter; sparse spatial sampling, ~3 day sub-satellite repeat |
| Minimum detectable regional flux anomaly | Seasonal aggregation required; single-overpass anomalies below ~5 ppb typically not separable from noise without multi-week averaging |
| Archive depth | Sentinel-5P: from May 2018; Sentinel-1: from April 2014; MODIS: from 2000 |
| Delivery formats | GeoTIFF inundation maps, NetCDF column anomaly grids, CSV seasonal flux tables with uncertainty ranges |
Analytics Satellize can run
| Seasonal paddy inundation calendar | Sentinel-1 SAR backscatter time-series classification; threshold or random-forest approach on VV/VH change detection | GeoTIFF stack of weekly inundation extent per field season, with transplanting and drainage date rasters |
| TROPOMI XCH4 regional anomaly time series | Quality-filtered TROPOMI L2 column aggregation; background subtraction using upwind reference pixels; cloud-gap flagging | Monthly NetCDF anomaly grids with observation-count and cloud-fraction layers; PDF seasonal summary report |
| Paddy-area-weighted emission prior map | Overlay of SAR inundation extent with published per-hectare emission factor ranges (IPCC Tier 2 or regional studies); spatial allocation to TROPOMI grid | NetCDF emission prior at 0.1-degree resolution for input to client transport model |
| Source-separation uncertainty assessment | Co-location of wetland extent (from published Global Surface Water or similar open datasets), livestock density grids, and fire radiative power from MODIS to quantify non-paddy contributions within each TROPOMI pixel | Tabular breakdown of estimated source fractions per region with confidence intervals; flagged pixels where source ambiguity exceeds defined threshold |
| Inventory cross-check report | Comparison of satellite-derived seasonal flux estimate against EDGAR or national inventory figures; statistical significance testing of discrepancies | PDF report suitable for inclusion in national GHG inventory documentation; flagged regions for ground-truth follow-up |
| Water management intervention signal detection | SAR-based identification of mid-season drainage events; correlation with TROPOMI column reduction in subsequent overpasses | GIS layer of fields with detected drainage events; associated column anomaly difference map |
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.