Open biomass burning emission-factor mapping for offset baselines
Savanna and crop-residue fires release vastly different quantities of CO₂, CH₄ and N₂O depending on fuel moisture, vegetation type and combustion completeness. Satellite-derived fire radiative power and fuel-load estimates let project developers move beyond IPCC Tier 1 defaults and build spatially explicit, defensible emission-factor baselines.
Sensors
- MODIS Terra/Aqua (MCD14ML fire product): Detects active fires at 1 km thermal resolution with twice-daily combined overpass frequency. Fire Radiative Power (FRP) in MW is the key output; the public MCD14ML collection provides per-detection FRP values used to estimate fire radiative energy and, from that, fuel consumption.
- VIIRS S-NPP / NOAA-20 (VNP14 / VJ114 fire products): 375 m active-fire detection with FRP retrieval, roughly four combined daily overpasses at mid-latitudes. The finer pixel size resolves fire edges that MODIS smears, reducing the saturation problem for intense agricultural burns. NOAA STAR distributes near-real-time products.
- Sentinel-2 MSI: 10 m optical imagery in visible and near-infrared, 20 m in shortwave infrared. Used to map pre-fire vegetation type and derive the Normalised Difference Vegetation Index and Normalised Burn Ratio. Revisit is 5 days at the equator (2-satellite constellation). Cloud cover is the primary constraint in humid tropics.
- Sentinel-1 SAR (C-band, IW mode): 10 m ground range resolution, 6-day repeat (combined A/B). C-band backscatter correlates with above-ground grass and shrub biomass up to roughly 1–2 kg m⁻², beyond which the signal saturates. Useful for estimating available fuel load in open savannas; less reliable in dense woodland.
Why default emission factors cost offset projects money
The IPCC Tier 1 approach assigns a single emission factor to broad land-cover classes. For savanna fire, the default combustion factor is 0.45 and the emission factor for methane is around 2.3 g kg⁻¹ dry matter burned, with N₂O at roughly 0.21 g kg⁻¹. These figures are continental averages assembled from field campaigns spanning decades and multiple vegetation formations. Applying them to a specific project area in, say, northern Australia or the West African Sahel can introduce errors of 30–60 percent in either direction, depending on actual fuel loads and fire behaviour.
That uncertainty is not merely academic. Fire-management offset methodologies, including the well-documented savanna-burning programmes operating under Australian Carbon Credit Unit rules, require proponents to demonstrate that their managed-fire regime produces fewer emissions than the unmanaged baseline. If the baseline emission factor is wrong, the credit quantity is wrong. Satellite-derived Tier 2 factors do not eliminate uncertainty, but they localise it and make it auditable.
From fire radiative power to fuel consumption: the physics
Fire Radiative Power is the instantaneous rate of radiative energy release from a fire, measured in megawatts per MODIS or VIIRS pixel. The link to fuel consumption rests on a radiative energy balance: integrating FRP over the fire's duration gives Fire Radiative Energy (FRE), and published coefficients, derived from controlled burns and field campaigns, convert FRE to dry-matter combusted. The coefficient most widely cited in the peer-reviewed literature sits around 0.368 kg MJ⁻¹, though it carries a coefficient of variation of roughly 11 percent even under controlled conditions.
The practical problem is temporal sampling. MODIS and VIIRS each pass over a given point roughly twice per day. A fast-moving grass fire may ignite, burn and extinguish between overpasses. Studies comparing satellite FRE with aircraft-measured fuel consumption find that polar-orbiting sensors capture somewhere between 20 and 80 percent of total FRE depending on fire duration and local overpass timing. Geostationary sensors (Meteosat SEVIRI, Himawari AHI) fill the temporal gap but at coarser spatial resolution, around 3–4 km. A credible Tier 2 workflow uses polar sensors for FRP magnitude and geostationary sensors for temporal completeness, then applies a correction factor derived from the local fire-duration climatology.
Fuel load: what SAR and optical indices actually resolve
Emission quantity is FRP-derived fuel consumption multiplied by an emission factor per unit mass. That emission factor itself varies with fuel moisture and vegetation type, so the pre-fire fuel characterisation matters. Sentinel-2 NDVI and the Normalised Difference Moisture Index (NDMI, using bands 8A and 11) provide a proxy for canopy greenness and moisture at 20 m. Dry-season NDVI in savanna correlates reasonably well with standing herbaceous biomass up to around 400 g m⁻², beyond which the relationship flattens.
Sentinel-1 C-band SAR adds a structural dimension. Backscatter in VV and VH polarisations responds to canopy water content and surface roughness. In open grassland, the correlation with fuel load holds below roughly 1–2 kg m⁻² of above-ground dry matter. Woodland understory is harder: the canopy attenuates the signal and the relationship becomes ambiguous. L-band SAR (ALOS-2 PALSAR-2, or the forthcoming NISAR) penetrates further, but those data are not freely available at Sentinel cadences. The honest position is that satellite-derived fuel loads carry uncertainties of 20–40 percent in heterogeneous savannas, and field calibration plots remain necessary for Tier 2 compliance.
Non-CO₂ species: where the uncertainty is largest and most consequential
CO₂ from complete combustion is relatively well constrained. The non-CO₂ species are not. Methane, nitrous oxide, carbon monoxide and non-methane volatile organic compounds are all emitted in quantities that depend heavily on combustion efficiency, which correlates with fire intensity and fuel moisture. The IPCC emission factor for CH₄ from savanna burning has an uncertainty range of roughly plus or minus 50 percent at the 95 percent confidence level. N₂O is worse.
Satellite fire radiative power provides a proxy for combustion efficiency: high FRP per unit area generally indicates flaming-dominated combustion with lower CH₄ and CO yields relative to CO₂, while low FRP suggests smouldering with higher ratios. This relationship is supported by airborne campaign data but is not yet embedded in any standard MRV protocol as a mandatory correction. Using it requires project developers to document the method transparently and accept that verification bodies may apply conservative adjustments. Satellize's view, for what it is worth, is that a documented Tier 2 uncertainty range of 40 percent is more credible to a sophisticated buyer of offsets than a Tier 1 point estimate with hidden error.
Building the spatially explicit baseline: workflow in outline
A practical workflow has four stages. First, map the pre-fire fuel landscape using multi-date Sentinel-2 composites and Sentinel-1 mosaics to classify vegetation type and estimate above-ground biomass by stratum. Second, extract historical fire perimeters from the MODIS MCD64A1 burned-area product (500 m, monthly) or the higher-resolution VIIRS VNP64A1 product, going back as far as the 2000 archive allows for MODIS or 2012 for VIIRS. Third, pair each historical burn event with its FRP time series from MCD14ML or VNP14, integrate to FRE, and convert to fuel consumption using the published coefficient. Fourth, apply stratum-specific emission factors from the IPCC Tier 2 tables, adjusted where combustion-efficiency proxies from FRP support a departure from the default.
The output is a gridded map of historical emission intensity (tonnes CO₂-equivalent per hectare per fire event) across the project area and its buffer zone. That map becomes the baseline against which managed-fire outcomes are measured. Satellize has built analogous spatially explicit baselines for agricultural contexts, including the Tonga crop-estimation programme, and the underlying data-fusion logic transfers directly to savanna fire accounting. The archive depth of MODIS (2000 to present) is particularly valuable: twenty-plus years of fire history is long enough to capture the interannual variability that a single-year baseline would miss.
Honest limits and what they mean for project design
Three constraints deserve explicit acknowledgement before a project developer commits to this approach. Cloud cover interrupts optical fuel-load mapping during the wet season, exactly when green-up determines the following dry-season fuel load. SAR partially compensates but does not fully substitute. Second, the FRP-to-FRE conversion assumes that the fire's radiative fraction is roughly constant, an assumption that breaks down for very intense fires where convective energy dominates. Third, MODIS at 1 km and even VIIRS at 375 m will miss small patchy burns below roughly one to two pixels in extent. In a landscape with many small agricultural fires, cumulative emissions from sub-pixel burns can be material.
None of these limits disqualify the satellite-based Tier 2 approach. They do mean that the uncertainty bounds on the final emission factor should be reported honestly, that field validation campaigns at the start of a project are not optional, and that the methodology document should specify how sub-pixel fire omission is handled. A registry-grade evidence package built on these data is achievable; a registry-grade evidence package with implausibly tight confidence intervals is not.
Typical figures
| MODIS FRP spatial resolution | 1 km per pixel (MCD14ML active fire product) |
| VIIRS FRP spatial resolution | 375 m per pixel (VNP14 / VJ114 products) |
| Combined MODIS Terra + Aqua overpass frequency | ~4 times per day globally; ~2 times per day per sensor |
| Sentinel-1 SAR resolution and revisit | 10 m ground range (IW mode); 6-day repeat with A+B constellation |
| Sentinel-2 MSI resolution and revisit | 10 m (visible/NIR), 20 m (SWIR); 5-day equatorial revisit (A+B) |
| MODIS burned-area archive depth | November 2000 to present (MCD64A1 monthly, 500 m) |
| FRE-to-fuel-consumption coefficient | ~0.368 kg MJ⁻¹ (published range; CV ~11% under controlled conditions) |
| Minimum detectable fire size (VIIRS) | ~0.1 ha under favourable conditions; sub-pixel fires not detected |
| Sentinel-1 fuel-load saturation threshold | ~1–2 kg m⁻² above-ground dry matter (C-band) |
| Delivery formats | GeoTIFF emission-factor grids, GeoPackage fire-history polygons, CSV per-event FRE tables, PDF methodology report |
Analytics Satellize can run
| Pre-fire fuel-load map by vegetation stratum | Sentinel-2 NDVI/NDMI time-series compositing combined with Sentinel-1 VV/VH backscatter regression against published biomass field data | GeoTIFF raster (kg dry matter m⁻²) with stratum classification layer and per-class uncertainty bounds |
| Historical fire perimeter and burn-severity archive | MODIS MCD64A1 burned-area product supplemented by Sentinel-2 dNBR (differenced Normalised Burn Ratio) for high-resolution edge delineation | GeoPackage polygon archive with date, area, mean dNBR and severity class per event; 2000-to-present for MODIS, 2017-to-present for Sentinel-2 |
| Per-event fire radiative energy and fuel-consumption estimate | MODIS MCD14ML and VIIRS VNP14 FRP time series integrated to FRE using published radiative energy balance coefficients; geostationary gap-fill where available | CSV table of fire events with FRE (MJ), estimated fuel consumption (kg ha⁻¹) and temporal-sampling correction factor |
| Spatially explicit emission-factor baseline grid | IPCC Tier 2 framework: stratum-specific emission factors applied to fuel-consumption estimates, with combustion-efficiency adjustment derived from FRP intensity percentile | GeoTIFF grid of historical mean emission intensity (t CO₂-eq ha⁻¹ fire event⁻¹) with 90% confidence interval rasters |
| Non-CO₂ species emission ratio analysis | FRP-percentile-based combustion-efficiency proxy applied to IPCC Tier 2 emission-factor ranges for CH₄, N₂O and CO; uncertainty propagation following published Monte Carlo approaches | Per-species emission-factor report with documented uncertainty ranges, suitable for inclusion in a Verra or Gold Standard methodology annex |
| Annual baseline update and interannual variability assessment | Rolling update of fuel-load and fire-history layers; Mann-Kendall trend test on annual emission intensity to flag non-stationarity in the baseline period | Annual GIS layer refresh plus a one-page statistical summary flagging any trend that would require baseline revision under registry rules |
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.