Post-fire ash cover and nutrient deposition mapping on cropland
Stubble and field-margin fires deposit ash unevenly across cropland, creating patches of elevated potassium and calcium that standard soil sampling misses. Sentinel-2 and Landsat 8/9 resolve the spectral contrast between ash, char, and bare soil well enough to map that variability at field scale.
Sensors
- Sentinel-2 MSI: 10 m visible and NIR bands, 20 m SWIR bands (1610 nm and 2190 nm); 5-day revisit at the equator with both satellites. The SWIR-to-red ratio is the primary discriminator between high-albedo ash and dark char. Cloud cover is the main operational constraint.
- Landsat 8/9 OLI: 30 m multispectral including SWIR1 (1565–1651 nm) and SWIR2 (2107–2294 nm); 16-day single-satellite revisit, 8-day combined. Longer archive (Landsat 8 from 2013, Landsat 9 from 2021) supports multi-season change analysis. Coarser than Sentinel-2 but consistent radiometric calibration aids time-series work.
- VIIRS 375 m active-fire product (VNP14IMG): Near-real-time fire detections at 375 m pixel resolution using mid-infrared and thermal infrared bands. Anchors the fire timing for pre/post compositing. Small field fires below roughly one to two hectares are frequently missed at this resolution; the detection floor is a known limitation in fragmented agricultural landscapes.
- MODIS MCD64A1 burned-area product: Monthly global burned-area at 500 m, derived from MODIS surface reflectance time series. Useful for regional-scale audit of fire frequency and seasonality on cropland, but too coarse to resolve individual field boundaries or intra-field ash distribution.
What ash actually looks like from orbit
Fresh wood and crop ash has a high visible albedo, often appearing pale grey to near-white in true-colour imagery, while charred residue is strongly absorbing across visible and near-infrared wavelengths. Bare mineral soil sits spectrally between the two. This three-way contrast is clearest in the shortwave infrared: ash reflectance in the 1600 nm and 2200 nm bands remains relatively high, whereas char drops sharply. The ratio of SWIR2 to red reflectance, sometimes formalised as a Char Index, exploits this separation. It is not a perfect discriminant. Dry sandy soils and some carbonate-rich surfaces can mimic ash albedo, which is why a pre-fire baseline image is necessary rather than optional.
The normalised burn ratio (NBR), computed from NIR and SWIR2, was designed for forest fire severity mapping, but its differenced form (dNBR, comparing pre- and post-fire images) transfers reasonably well to cropland when the pre-fire surface is vegetated stubble rather than bare soil. The complication is that agricultural land cycles through bare-soil states routinely, so a dNBR computed without careful pre-fire image selection can conflate tillage with burning. Anchoring the analysis to VIIRS or MODIS active-fire detections, even imperfect ones, constrains the timing and reduces false positives.
From spectral signal to nutrient proxy
Ash depth is the agronomically relevant variable. Published field studies report that crop residue fires typically produce ash layers ranging from a few millimetres to around two centimetres depending on biomass density and combustion completeness. Potassium and calcium concentrations in cereal ash are well documented in the agronomic literature: wheat straw ash, for instance, carries potassium content in the range of 5 to 15 percent by dry weight depending on variety and soil conditions. The satellite cannot measure depth or chemistry directly. What it measures is the fractional cover of ash across a pixel, which, combined with a biomass estimate from a pre-fire vegetation index, provides a proxy for total deposited material.
The proxy is coarse. Combustion completeness varies with wind speed, moisture content, and fire intensity, none of which are resolved at Sentinel-2 scale. Ash redistribution by wind between the fire event and the satellite overpass introduces further uncertainty. The honest use of the satellite product is to map spatial variability in deposition, not to replace soil sampling. A field that shows 60 percent ash cover in one corner and 10 percent in another gives a precision-agriculture prescription zone that would be invisible to a standard five-point composite soil sample.
Timing is almost everything
Ash weathers rapidly. Rain, wind, and tillage can reduce spectral contrast to background levels within days to weeks of a fire. The operational window for useful imagery is short: ideally one to five days post-fire, before the first significant rainfall event. This makes cloud cover the dominant practical constraint in humid temperate regions. In semi-arid agricultural zones, the window is longer but wind redistribution is faster.
VIIRS active-fire detections, available through NASA FIRMS with latency of a few hours, provide the trigger for tasking or archive search. The 375 m detection floor means that fires confined to narrow field margins or small plots below roughly one to two hectares will not appear in the VIIRS record. In those cases, the post-fire spectral anomaly in Sentinel-2 may be the only evidence the fire occurred at all. Conversely, VIIRS detections in areas with no corresponding Sentinel-2 spectral change are usually attributable to fires that were extinguished before significant ash accumulated, or to detection artefacts near industrial heat sources.
Resolution limits and the small-field problem
Sentinel-2's 10 m visible bands and 20 m SWIR bands are well matched to fields above roughly one hectare. Below that, mixed pixels dilute the ash signal against surrounding unburned soil, and the Char Index value becomes unreliable. Landsat's 30 m SWIR bands are worse in this respect: a 30 m pixel over a fragmented smallholder landscape may contain portions of three or four different fields, making field-level attribution impossible without sub-pixel unmixing.
Commercial very-high-resolution imagery (sub-metre to 3 m) can resolve ash patches on individual beds within a field, but it lacks the systematic revisit needed to catch the short post-fire window reliably. The practical answer for smallholder contexts is to use Sentinel-2 for spatial extent mapping and flag fields that warrant ground verification, rather than to attempt per-field nutrient budgeting from satellite data alone. This is roughly the same logic Satellize applies in its Tonga crop-estimation programme, where open-constellation data sets the spatial framework and ground data provides the calibration.
What the output is actually good for
A post-fire ash-cover map at 10 to 20 m resolution, delivered within a week of the fire event, gives an agronomist or extension officer three things a field visit cannot easily provide: spatial extent of the burn, intra-field variability in ash deposition, and a documented baseline for comparing subsequent soil test results. For insurance purposes, it provides an independent, time-stamped record of fire occurrence and affected area, distinct from farmer self-reporting.
For regional agricultural agencies, multi-season fire frequency maps derived from MODIS MCD64A1 or annual Sentinel-2 change composites identify which districts are burning residue systematically. That has policy relevance: repeated burning depletes soil organic carbon over time, even as it deposits short-term mineral nutrients. The satellite record going back to 2013 for Landsat 8 and 2015 for Sentinel-2 is long enough to characterise seasonal burning patterns and correlate them with yield trends from other sources, though establishing causality requires ground-truth data that satellite analysis alone cannot supply.
Typical figures
| Primary spatial resolution (SWIR bands) | 20 m (Sentinel-2 MSI); 30 m (Landsat 8/9 OLI) |
| Primary spatial resolution (visible/NIR) | 10 m (Sentinel-2 MSI); 30 m (Landsat 8/9 OLI) |
| Revisit frequency | 5 days at equator, both Sentinel-2 satellites combined; 8 days combined Landsat 8+9 |
| Fire-detection latency (VIIRS FIRMS) | Typically 3 to 6 hours from overpass |
| Minimum detectable fire (VIIRS 375 m) | Approximately 1 to 2 ha for field fires; smaller fires frequently missed |
| Key spectral bands for ash/char discrimination | SWIR1 (~1610 nm), SWIR2 (~2190 nm), Red (~665 nm), NIR (~842 nm) |
| Operational post-fire mapping window | 1 to 5 days post-fire (before rain or tillage degrades signal) |
| Archive depth | Sentinel-2 from 2015; Landsat 8 from 2013; MODIS burned area from 2000 |
| Burned-area product resolution (MODIS MCD64A1) | 500 m monthly; insufficient for individual field mapping |
| Typical deliverable format | GeoTIFF raster (ash-cover fraction, dNBR, Char Index); field-boundary polygon layer with per-field statistics |
Analytics Satellize can run
| Post-fire ash-cover fraction map | Spectral unmixing using SWIR/red Char Index on Sentinel-2 20 m imagery, calibrated against pre-fire bare-soil reflectance | GeoTIFF raster at 20 m, clipped to agricultural land mask, with per-field mean and variance statistics in GeoPackage |
| dNBR severity classification for cropland | Differenced Normalised Burn Ratio (pre- minus post-fire NBR) on Sentinel-2 or Landsat 8/9, with fire timing anchored to VIIRS VNP14IMG detections | Classified raster (unburned, low, moderate, high severity) with area statistics per administrative unit, PDF report |
| Fire-event timing and extent record | VIIRS 375 m active-fire point detections cross-referenced with Sentinel-2 spectral change; MODIS MCD64A1 for regional context | Timestamped polygon layer of confirmed burn extents, suitable for insurance or regulatory audit, delivered as GeoJSON |
| Intra-field ash deposition variability zones | K-means or threshold-based zoning of ash-cover fraction within field boundaries, generating variable-rate application zones | Prescription-zone shapefile compatible with common farm-management software, with recommended soil-sampling locations flagged |
| Multi-season fire frequency map | Annual burned-area compositing from Sentinel-2 dNBR time series (2015 to present) or MODIS MCD64A1 (2000 to present) over agricultural land mask | Raster showing fire return interval per pixel, exported as GeoTIFF with summary statistics by district or farm unit |
| Potassium deposition proxy estimate | Ash-cover fraction combined with pre-fire above-ground biomass proxy (NDVI-derived) and published ash chemistry ranges for the relevant crop type; output is an order-of-magnitude estimate with explicit uncertainty bounds | Tabular report per field with estimated K deposition range (kg/ha), confidence class, and recommendation to verify with soil sampling |
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.