Rice paddy methane emission monitoring for agricultural offset projects
Flooded rice paddies are a major agricultural methane source. Alternate wetting and drying offset projects claim emission cuts that must be independently verified. SAR flood-cycle mapping, satellite crop calendars and TROPOMI atmospheric retrievals together form a defensible evidence chain.
Sensors
- Sentinel-1 SAR (C-band, ESA): 10 m ground resolution in Interferometric Wide Swath mode, 6-day repeat at equatorial latitudes with both satellites active. C-band backscatter distinguishes open water from vegetated or bare soil surfaces; temporal coherence change detects flooding onset and drainage events even through cloud and at night.
- ALOS-2 PALSAR-2 (L-band, JAXA): L-band (1.27 GHz) penetrates rice canopy more deeply than C-band, allowing flood detection beneath a partial crop cover that would saturate a C-band signal. Spatial resolution 3–10 m depending on mode; revisit approximately 14 days, limiting temporal precision but improving detection under dense canopy.
- Sentinel-5P TROPOMI: Methane column retrievals (CH4 total column) at 5.5 × 7 km pixel footprint, daily global coverage. Precision approximately 0.6–1% per sounding under clear-sky conditions. Used as a regional consistency check on claimed emission reductions, not as a field-scale measurement.
- Sentinel-2 MSI: 10–20 m multispectral imagery, 5-day revisit with both satellites. Near-infrared and red-edge bands track transplanting date, canopy greenness and harvest timing, anchoring the crop calendar that bounds the emission-reduction period claimed by a project.
Why water management is the whole argument
Methane from rice paddies is produced by methanogenic archaea in anaerobic waterlogged soil. Remove the standing water, even briefly, and oxygen re-enters the soil column, suppressing methanogenesis. Alternate wetting and drying (AWD) exploits this: farmers allow the field to dry to a threshold soil-water potential before re-flooding. Published field trials, including work coordinated under the GWP100 accounting frameworks used by Verra's VM0041 methodology, indicate AWD can reduce paddy methane emissions by 30–70% relative to continuously flooded baselines, depending on soil type, climate and the drying depth achieved.
The verification problem is that the emission reduction is entirely contingent on the water management actually happening as described. A project developer who claims AWD but continues conventional flooding earns carbon credits for emissions that were never avoided. Satellite observation does not measure methane flux directly at field scale; instead it verifies the hydrological precondition. If the precondition is absent, the claimed reduction is indefensible.
What SAR backscatter reveals about a flooded field
Open water returns very low C-band backscatter to the radar because a specular surface deflects the signal away from the sensor. Bare moist soil returns moderate backscatter. A flooded field with an emerging rice canopy produces a distinctive double-bounce signal as the radar pulse reflects off the water surface and then off vertical plant stems, increasing backscatter above both open-water and dry-soil values. These signatures are separable in Sentinel-1 VV and VH polarisation time series, and the transition from low to moderate backscatter marks drainage events with a timing precision of roughly 6 days given the constellation's revisit.
In practice, analysts construct a per-field time series of backscatter across the growing season and compare it against the flood/drain schedule the project developer has submitted. A field that was supposed to be drained for ten days in mid-tillering but shows continuously low backscatter throughout that window has not been managed as claimed. The method is well established in the literature; JAXA has used PALSAR time series for paddy mapping across Asia at national scales, and the Copernicus Global Land Service uses Sentinel-1 for surface-water dynamics at 10 m. Cloud cover is not an obstacle for SAR, which is the principal reason radar is preferred over optical for this application in tropical and monsoonal environments.
Anchoring the crop calendar with Sentinel-2
Methane emission rates vary sharply with crop growth stage. Emissions peak during the mid-vegetative and reproductive phases and fall after drainage prior to harvest. A project that claims AWD during the wrong growth window, or that misreports transplanting and harvest dates to extend the crediting period, can inflate the claimed reduction. Sentinel-2 time series of the normalised difference vegetation index and the red-edge chlorophyll index resolve transplanting to within roughly 10 days and heading date to similar precision in cloud-permitting conditions.
The crop calendar derived from Sentinel-2 is used to constrain the emission model, not to replace field measurement. If the satellite-derived transplanting date is three weeks later than the project report states, the crediting period is correspondingly shorter, and any AWD events claimed before transplanting are irrelevant to methane flux. This cross-check is simple to automate across large project areas and catches reporting errors that ground audits, conducted at single points in time, routinely miss.
TROPOMI as a regional sanity check, and its limits
TROPOMI measures the total atmospheric column of methane at a footprint of 5.5 × 7 km. A single paddy field of a few hectares is invisible to it. But a rice-growing basin of several hundred thousand hectares, if genuinely managed under AWD at scale, should produce a detectable suppression of the seasonal methane anomaly relative to years without AWD adoption, assuming atmospheric conditions permit retrieval and the signal is not swamped by other regional sources such as livestock, wetlands or fossil-fuel infrastructure.
This is a consistency check, not a verification tool. If TROPOMI shows no suppression in a region where a large portfolio of AWD projects claims significant aggregate reductions, that is a flag worth investigating. It does not prove fraud, because transport modelling uncertainty and source mixing can obscure real reductions at this scale. Conversely, a TROPOMI signal consistent with claimed reductions does not confirm them at field level. The honest position is that atmospheric retrievals bound the plausible range; SAR and crop-calendar evidence do the field-level work.
Honest limits of the satellite evidence chain
SAR flood detection has a spatial resolution floor. Sentinel-1 at 10 m can resolve individual paddy plots in most Asian rice landscapes, but very small or irregularly shaped fields may be mixed with adjacent land cover, degrading classification accuracy. PALSAR-2's L-band advantage under dense canopy comes at a cost: 14-day revisit means a short AWD event of five to seven days can fall entirely between acquisitions and go undetected. Neither sensor measures soil-water potential, which is the agronomic variable that actually controls methanogenesis; they measure surface inundation state, which is a proxy.
Sentinel-2 cloud contamination in monsoonal growing seasons can leave gaps of two to four weeks in the optical record, reducing crop-calendar precision precisely when it matters most. Fusion with SAR-derived phenology metrics partially compensates. Finally, the satellite record verifies hydrological practice but does not replace emission-factor measurement. The actual methane reduction per unit of drained time depends on soil organic carbon content, temperature and fertiliser regime, none of which are observable from orbit. Satellite evidence is a necessary component of AWD MRV, not a sufficient one. Satellize's crop-estimation work for the Kingdom of Tonga used a similar fusion of optical and ancillary data to bound agronomic uncertainty, and the same principle applies here.
Building an audit-grade evidence package
A credible MRV submission for an AWD offset project needs, at minimum: a per-field flood-state time series covering the full crediting period, derived from Sentinel-1 with PALSAR-2 gap-filling where available; a satellite-derived crop calendar with transplanting and harvest dates; a comparison of observed flood/drain events against the project's submitted water-management schedule; and a TROPOMI regional methane anomaly assessment for the basin. Each layer should carry explicit uncertainty bounds and flag periods where cloud or revisit gaps reduce confidence.
Verifiers reviewing such a package can identify fields where SAR evidence is inconclusive and prioritise those for ground truthing, rather than sampling randomly across a project area. That targeted approach is more efficient and more defensible than either pure satellite or pure ground-based audit. The archive depth of Sentinel-1 (operational since 2014 for Sentinel-1A, 2016 for Sentinel-1B) and Sentinel-2 (2015 onwards) means historical baselines can be reconstructed for projects seeking retrospective crediting, subject to the methodology's rules on baseline periods.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | 10 m (range) × 10 m (azimuth) after multi-look processing |
| SAR revisit (Sentinel-1, both satellites) | 6 days at equatorial latitudes; up to 12 days with one satellite |
| L-band SAR revisit (PALSAR-2) | Approximately 14 days; 3–10 m resolution depending on observation mode |
| Optical resolution (Sentinel-2 MSI) | 10 m (visible/NIR bands), 20 m (red-edge and SWIR bands) |
| Optical revisit (Sentinel-2, both satellites) | 5 days at equatorial latitudes under cloud-free conditions |
| TROPOMI CH4 pixel footprint | 5.5 × 7 km; daily global coverage |
| TROPOMI CH4 retrieval precision | Approximately 0.6–1% per sounding (clear-sky); cloud-contaminated scenes excluded |
| Minimum detectable AWD drainage event (SAR) | Events shorter than the revisit interval (< 6 days for Sentinel-1) may be missed; L-band gap-filling recommended |
| Sentinel-1 archive depth | Sentinel-1A from April 2014; Sentinel-1B from April 2016 (note: Sentinel-1B lost in 2021, Sentinel-1C launched 2024) |
| Sentinel-2 archive depth | Sentinel-2A from June 2015; Sentinel-2B from March 2017 |
Analytics Satellize can run
| Per-field flood-state time series | Sentinel-1 VV/VH backscatter thresholding and change detection; PALSAR-2 L-band integration for canopy-obscured periods | GIS polygon layer with flood/drain state per field per acquisition date, flagged against project-submitted water-management schedule |
| Satellite-derived crop calendar | Sentinel-2 NDVI and red-edge index time-series phenology fitting to extract transplanting, heading and harvest dates | Tabular crop-calendar report per project parcel with date estimates and confidence intervals; comparison against project-reported dates |
| AWD compliance score per field | Comparison of observed drainage events (SAR) against claimed AWD schedule within the satellite-derived crediting window (crop calendar) | Field-level compliance matrix showing confirmed, unconfirmed and contradicted AWD events; priority list for ground-truthing |
| Regional methane anomaly assessment | TROPOMI CH4 column time-series analysis for the project basin, referenced against multi-year seasonal baseline; simple atmospheric transport screening | Basin-scale methane anomaly report with seasonal plots; qualitative consistency rating against aggregate project claims |
| Historical baseline flood-cycle reconstruction | Sentinel-1 and Sentinel-2 archive analysis for pre-project years to characterise conventional flooding practice in the project area | Multi-year baseline flood-state dataset supporting additionality and baseline emission-factor calculations under Verra VM0041 or equivalent methodology |
| Audit-grade evidence package | Integration of all above layers with uncertainty quantification, gap-flagging and metadata compliant with third-party verifier requirements | Structured PDF and GIS archive suitable for submission to a validation and verification body; includes data-quality flags and recommended ground-audit priority fields |
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.