Flooded rice paddy mapping for methane emission inventories
Flooded paddy fields are the largest single agricultural source of methane. Combining Sentinel-1 radar backscatter with Sentinel-2 optical phenology lets analysts map inundation extent and flood duration at field scale, the two variables that drive UNFCCC emission factor calculations.
Sensors
- Sentinel-1 C-band SAR (ESA): 5 by 20 metre resolution in Interferometric Wide Swath mode, 6-day repeat at mid-latitudes with the two-satellite constellation. C-band (5.405 GHz) penetrates cloud and light rain. Standing water under a sparse early-season canopy produces strong specular reflection (low backscatter), while a closed canopy over flooded soil generates double-bounce return. Tracking the transition between these two signatures through the growing season is the core detection mechanism.
- Sentinel-2 MSI (ESA): 10 metre resolution in visible and near-infrared bands, 20 metre in red-edge and shortwave infrared. Five-day revisit with both satellites. NDVI and LSWI time series identify transplanting dates, heading, and senescence. Cloud cover is the primary limit; in humid tropical rice regions, 30 to 50 per cent of acquisitions may be unusable in the wet season, which is exactly when paddies are flooded.
- ALOS-2 PALSAR-2 L-band SAR (JAXA): L-band (1.27 GHz) penetrates a denser canopy than C-band, making it more sensitive to double-bounce from flooded soil under a closed rice canopy at later growth stages. Spatial resolution in stripmap mode reaches 3 metres, though standard archive products are typically 6 to 10 metres. Revisit is 14 days for a single satellite, which limits temporal density compared with Sentinel-1.
- Landsat 8 and 9 OLI (USGS/NASA): 30 metre resolution, 16-day revisit per satellite (8-day combined). Less useful than Sentinel-2 for phenology in cloudy regions, but the archive extends to 1972 across the Landsat programme, enabling multi-decadal trend analysis of paddy area change. SWIR bands distinguish flooded from non-flooded soil when the canopy is sparse.
Why flood duration, not just paddy area, determines the emission
The IPCC methodology for estimating methane from rice cultivation (Tier 1 and Tier 2 in the 2006 Guidelines) multiplies a baseline emission factor by a scaling factor for water regime and then by the number of cultivation days under flooded conditions. Paddy area alone is insufficient. A field flooded for 120 days emits roughly twice as much methane as the same field flooded for 60 days, all else equal. National inventories that rely solely on agricultural census data for paddy area, without any satellite-derived flood duration, are making a structural assumption that is rarely validated against actual field conditions.
Methane is produced by methanogenic archaea in anaerobic waterlogged soil. Drainage events, even brief mid-season ones, interrupt methanogenesis and can reduce seasonal emissions by 30 to 50 per cent according to published field studies. Satellite time series can detect those drainage windows. That distinction matters enormously for countries trying to claim emission reductions under UNFCCC frameworks, or for carbon market methodologies that reward improved water management.
What a flooded field looks like to a radar
C-band SAR backscatter from a flooded paddy passes through three recognisable phases. At transplanting, when seedlings are a few centimetres tall and water is clearly visible, the surface acts as a near-specular reflector. Backscatter in VV polarisation drops sharply, typically below minus 15 dB in Sentinel-1 data, distinguishing flooded paddies from dry soil or upland crops. As the canopy closes, volume scattering from the vegetation increases and backscatter rises. Later, when the crop is mature and the field is drained before harvest, backscatter rises further and the specular signature disappears entirely.
Mapping flood duration means fitting this temporal backscatter trajectory to each pixel across a growing season. The practical challenge is that the trajectory is noisy: wind roughens the water surface and raises apparent backscatter; heavy rainfall does the same. Analysts typically apply a threshold on a multi-temporal VV or VH stack, combined with a phenological constraint from optical data to exclude non-rice inundated surfaces such as reservoirs or wetlands. L-band from PALSAR-2 adds sensitivity at the canopy-closure stage where C-band begins to saturate, though the 14-day revisit means fewer temporal samples per season.
Fusing SAR and optical phenology: where the two methods depend on each other
Neither sensor family solves the problem alone. SAR detects water but cannot reliably distinguish rice from other flooded vegetation without phenological context. Optical imagery provides that context through the NDVI trajectory, which for transplanted rice shows a characteristic rise from near zero at transplanting to a peak around heading, then a rapid decline at senescence. The LSWI (Land Surface Water Index, derived from Sentinel-2 NIR and SWIR bands) is particularly sensitive to canopy water content and helps confirm the flooded condition early in the season.
The fusion workflow typically proceeds in two passes. First, optical time series define the probable transplanting window and crop calendar for each pixel, even if some dates are cloud-contaminated and must be interpolated. Second, SAR backscatter within that phenological window is used to estimate the actual inundation period day by day. The result is a per-pixel flood duration estimate in days, which feeds directly into the IPCC scaling factor calculation. Uncertainty is highest in regions with persistent cloud cover and short growing seasons, where the SAR record may contain only three or four cloud-free-equivalent acquisitions per phase.
Honest limits: what the method cannot tell you
At Sentinel-1's 20-metre posting in IW mode, a paddy bund or a narrow drainage channel is sub-pixel. Small fields below roughly 0.1 hectares are likely to be mixed with adjacent land cover, which biases flood duration estimates downward in fragmented smallholder landscapes. This is a real problem across much of South and Southeast Asia, where average field sizes in countries such as Bangladesh and the Philippines are well under 0.5 hectares.
The method also cannot distinguish between flooded fallow and flooded rice in the early season without optical confirmation. Aquaculture ponds, which are expanding rapidly across the Mekong Delta and coastal Southeast Asia, can produce SAR signatures similar to early-season paddies. Misclassification inflates paddy area estimates. Careful masking using cadastral data or higher-resolution commercial imagery reduces this error but adds cost and complexity. Finally, the approach gives inundation extent, not water depth. Shallow flooding below a few centimetres may not sustain full methanogenesis, but the satellite cannot measure depth; that requires ground truth or hydrological modelling.
From pixel counts to a national inventory line
The output of the remote sensing analysis is a set of rasters: paddy extent per season, flood onset date, flood end date, and derived flood duration in days. Converting these to an emission estimate requires applying the IPCC Tier 2 emission factors, which differ by water regime category (continuously flooded, intermittently flooded with single or multiple drainage events, and rainfed). The satellite record can distinguish continuously flooded from intermittently flooded fields if the revisit frequency is sufficient to detect drainage windows of five days or more. At six-day Sentinel-1 revisit, a three-day drainage event will be missed roughly half the time depending on acquisition geometry.
Countries submitting national greenhouse-gas inventories to the UNFCCC are increasingly expected to show how their paddy area estimates were derived and to quantify uncertainty. Satellite-derived estimates, when documented with a clear processing chain, provide an auditable and reproducible basis that census-based figures cannot match. Satellize has built comparable phenology-driven area estimation pipelines for clients including the Kingdom of Tonga crop-estimation programme, and the same fusion architecture applies directly to paddy mapping at national scale. The analytic outputs are delivered as GIS layers and summary tables structured for direct input to IPCC inventory worksheets.
Archive depth and what it enables for trend analysis
Sentinel-1 data are available globally from 2014 (Sentinel-1A launch) and from 2016 with improved revisit after Sentinel-1B joined the constellation. Sentinel-1C launched in 2024, maintaining continuity. The ALOS PALSAR archive extends to 2006, and the original ERS-1 and ERS-2 C-band SAR missions provide fragmentary coverage back to the early 1990s. Together, these archives allow analysts to reconstruct paddy area and approximate flood duration for roughly a decade at consistent quality, and longer with reduced temporal density.
That archive depth matters for two reasons. First, it allows countries to back-calculate historical inventories on a consistent methodological basis, resolving discontinuities that arise when census methodology changes. Second, it enables detection of structural shifts: conversion of paddies to aquaculture, adoption of alternate wetting and drying irrigation, or abandonment of marginal paddy land. All of these changes affect national methane budgets and are difficult to detect from ground surveys alone at the speed at which land use is changing across Asia.
Typical figures
| Spatial resolution (SAR, operational) | Sentinel-1 IW mode: 5 × 20 m (range × azimuth); ALOS-2 PALSAR-2 stripmap: 3 to 10 m |
| Spatial resolution (optical) | Sentinel-2 MSI: 10 m (VIS/NIR), 20 m (red-edge, SWIR); Landsat OLI: 30 m |
| Revisit frequency | Sentinel-1: 6 days at mid-latitudes (two satellites); ALOS-2: 14 days; Sentinel-2: 5 days (two satellites) |
| SAR frequency and polarisation | Sentinel-1: C-band 5.405 GHz, VV and VH; ALOS-2 PALSAR-2: L-band 1.27 GHz, HH, HV, full-pol |
| Minimum mappable field size (practical) | Approximately 0.1 ha at Sentinel-1 resolution; smaller fields subject to mixed-pixel bias |
| Minimum detectable flood duration window | Approximately 5 to 6 days at Sentinel-1 revisit; shorter drainage events likely missed |
| Archive depth | Sentinel-1: 2014 to present; ALOS PALSAR: 2006 to 2011; Landsat: 1972 to present |
| Cloud penetration | SAR: all-weather; Sentinel-2/Landsat: cloud-contaminated acquisitions must be excluded or gap-filled |
| Delivery formats | GeoTIFF rasters (flood extent, duration); GeoPackage or Shapefile (paddy boundaries); CSV summary tables structured for IPCC inventory worksheets |
| Latency (operational monitoring) | Sentinel-1 NRT products available within 1 to 3 hours of acquisition; Sentinel-2 L2A within 3 to 5 hours |
Analytics Satellize can run
| Seasonal paddy extent map | Supervised or threshold-based classification of multi-temporal Sentinel-1 VV/VH backscatter, constrained by Sentinel-2 NDVI phenology window | GeoTIFF raster and polygon layer of paddy fields per growing season, attributed with crop calendar dates |
| Per-pixel flood duration estimate | Day-count of inundation periods derived from SAR backscatter time series below empirically or physically derived thresholds, with drainage-event detection | GeoTIFF raster of flood duration in days per season; histogram summary by administrative unit |
| Water regime classification | Categorisation of pixels into IPCC water regime classes (continuously flooded, single drainage, multiple drainage, rainfed) based on backscatter trajectory shape | Classified raster and area-summary table formatted for IPCC Tier 2 inventory worksheet input |
| Transplanting and heading date map | NDVI and LSWI inflection-point detection on dense Sentinel-2 time series, with harmonic or Savitzky-Golay smoothing to handle cloud gaps | Raster layers of transplanting date and heading date (day-of-year); crop calendar validation report |
| Multi-year paddy area trend analysis | Consistent reprocessing of Sentinel-1 archive (2014 to present) with fixed classification parameters to produce comparable annual extent estimates | Time-series chart and tabular summary of paddy area by season and administrative unit; change-detection layer flagging conversion or abandonment |
| Methane emission proxy estimate | Application of IPCC 2006 Guidelines Tier 1 or Tier 2 emission factors to satellite-derived paddy area and flood duration by water regime class | Tabular emission estimate with uncertainty range, structured for national inventory submission; methodology note suitable for UNFCCC technical review |
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.