Sugarcane yield and sucrose content estimation

Canopy reflectance in the red-edge and shortwave-infrared bands tracks chlorophyll decline and fibre accumulation as sugarcane matures, giving mills and governments an independent estimate of stalk biomass and sucrose content weeks before harvest.

Sensors

Sentinel-2 MSI: 10 m visible and near-infrared bands, 20 m red-edge bands B5 (705 nm), B6 (740 nm), B7 (783 nm) and SWIR bands B11 (1610 nm), B12 (2190 nm). Five-day revisit at the equator with both satellites. Free archive from 2015. The red-edge triplet is the primary spectral region for tracking chlorophyll decline during sucrose accumulation.
PRISMA (ASI): Hyperspectral imager, 30 m spatial resolution, 239 contiguous bands from 400 to 2500 nm at roughly 10 nm sampling. Single satellite gives irregular revisit (days to weeks depending on tasking priority). Resolves narrow absorption features linked to water content, lignin and cellulose that broad-band sensors conflate, which matters when separating sucrose-related spectral shifts from senescent background material.
Landsat 8/9 OLI: 30 m multispectral, 16-day single-satellite revisit (8-day combined). SWIR1 (1570 nm) and SWIR2 (2110 nm) bands correlate with canopy water content and fibre fraction. No dedicated red-edge band, so sensitivity to early maturation signals is lower than Sentinel-2, but the archive back to 1972 (Landsat 1 MSS) and to 2013 (OLI) supports long-term yield trend analysis.
Planet SuperDove: 3 m resolution, near-daily revisit, eight bands including a red-edge band at 705 nm. Useful for within-field variability mapping and identifying lodged or stressed patches before they dilute mill-gate quality. Commercial licence required; no free archive.
MODIS / VIIRS: 250–500 m resolution, daily global coverage. Too coarse for individual field estimates in fragmented landscapes, but useful for regional-scale phenology tracking and anomaly detection across entire production zones in Brazil's Centre-South or South Africa's KwaZulu-Natal.

What the canopy is actually telling you

Sugarcane sucrose accumulation is not a discrete event; it is a gradual physiological shift that begins when the plant stops prioritising vegetative growth and starts translocating sugars into the stalk. During this phase, chlorophyll concentration in the upper canopy declines, cell-wall fibre fraction rises, and canopy water content falls. Each of these changes has a measurable spectral signature.

Chlorophyll absorbs strongly at red wavelengths (around 670 nm) and the transition from red to near-infrared reflectance, the red-edge slope, shifts towards shorter wavelengths as chlorophyll degrades. Sentinel-2 bands B5, B6 and B7 straddle this slope at 705, 740 and 783 nm, allowing indices such as the Red-Edge Chlorophyll Index (CIre) and the MERIS Terrestrial Chlorophyll Index (MTCI) to track the decline quantitatively. Brazilian research programmes, including work published through EMBRAPA and the University of São Paulo, have calibrated these relationships against destructive sampling across commercial varieties in the Cerrado and São Paulo state.

The SWIR bands add a second information layer. As fibre fraction rises and free water in the stalk decreases, SWIR reflectance increases. The Cellulose Absorption Index and similar SWIR-ratio metrics have been used in South African research (notably from the South African Sugarcane Research Institute, SASRI) to estimate fibre content independently of canopy architecture. PRISMA's contiguous coverage across the 400–2500 nm range allows both chlorophyll and fibre features to be retrieved simultaneously without the band-gap ambiguity that affects multispectral sensors.

From spectral index to tonnes per hectare

The translation from a spectral index to a stalk-biomass figure requires a calibration step, and this is where the method's honest limits begin. Canopy reflectance integrates signal from green leaves, senescent leaves, exposed soil between rows, and any residue from previous harvests. Row orientation relative to the sensor view angle, planting density, and the degree of lodging all alter the mixture. Soil-adjusted indices such as SAVI and its derivatives reduce but do not eliminate soil contamination, particularly in ratoon crops where canopy closure is incomplete early in the season.

Empirical regression models, whether ordinary least squares, random forest or Gaussian process regression, trained on local calibration data typically achieve R² values in the 0.7–0.85 range for total stalk biomass when red-edge indices are used as predictors. Sucrose content (expressed as commercial cane sugar, CCS, in Australian and South African industry parlance, or as ATR in Brazil) is harder to estimate remotely than biomass because it depends on temperature stress history, irrigation management and variety genetics in ways that are not fully captured by canopy reflectance alone. Published studies report correlations between CIre-derived chlorophyll estimates and CCS, but the relationships are weaker than for biomass and require local recalibration each season.

Physics-based radiative transfer models, particularly PROSAIL (a coupling of the PROSPECT leaf model and the SAIL canopy model), offer a path to more transferable estimates by inverting canopy reflectance to retrieve leaf chlorophyll content and leaf area index explicitly. This avoids the need to retrain an empirical model for each new geography, though the inversion is computationally heavier and requires assumptions about canopy structure that may not hold across all varieties.

Cloud cover and the harvest window problem

Sugarcane is predominantly grown in tropical and subtropical belts where cloud cover is persistent for much of the year. In Brazil's São Paulo state, the main harvest runs from April to November, which coincides with the dry season and relatively low cloud frequency. In South Africa's KwaZulu-Natal, harvest peaks between July and December. In both cases, the spectral maturation signal needs to be captured in a narrow window before the crop is cut.

Sentinel-2's five-day revisit at the equator sounds adequate, but after cloud screening, usable observations in humid tropical zones can fall to four or five clear acquisitions per season. This is enough to fit a phenological curve if the timing is favourable, but a single cloud event during the critical maturation window can create a gap that forces interpolation. Combining Sentinel-2 with Landsat 9 (which is offset by roughly eight days) improves the probability of at least one clear observation per fortnight. PRISMA's irregular revisit makes it unreliable as the primary time-series sensor; it is better used for targeted high-quality snapshots to validate or anchor the multispectral time series.

SAR sensors such as Sentinel-1 penetrate cloud but do not directly measure the optical properties linked to sucrose. SAR backscatter correlates with biomass and canopy water content, and fusion of SAR with optical data can fill cloud gaps for biomass estimation. That topic is covered in the SAR-optical yield fusion page in this library.

Mill scheduling is the actual use case

The practical output of this analysis is not a map for its own sake. Mills need to sequence harvesting across hundreds of contracted fields to maintain continuous throughput while maximising average CCS at intake. A field cut too early carries lower sucrose; one cut too late risks lodging, ratoon damage and quality loss from post-maturity fermentation. The scheduling problem is combinatorial and currently solved largely by agronomist experience and ground sampling, which is expensive and spatially sparse.

Satellite-derived maturity indices, updated every five to ten days across an entire supply zone, allow the mill to rank fields by estimated CCS trajectory and adjust the cut order dynamically. The economic value is not in the absolute accuracy of the CCS estimate but in the relative ranking: knowing that block A is two weeks ahead of block B is actionable even if the absolute CCS prediction carries a ±1.5 unit uncertainty.

Satellize's analytics approach for this use case follows the same logic applied in the Kingdom of Tonga crop-estimation programme: combine open-constellation time series with local calibration data to produce decision-relevant outputs rather than research-grade maps. For a mill or a national sugar authority, the deliverable is a ranked harvest-readiness layer updated on each clear overpass, with confidence intervals that reflect cloud-gap interpolation uncertainty honestly.

What the method cannot do

Remote sensing of sucrose content has genuine limits that should be stated plainly before a procurement decision is made. Variety effects are large: two fields with identical spectral signatures can differ by two or three CCS units if they are planted with varieties that have different sucrose-accumulation rates. Without variety maps (which can themselves be derived from satellite phenology, but imperfectly), the model will average across this variance.

Irrigation management introduces another confound. Irrigated cane maintains higher canopy water content and greener leaves later into the season than rainfed cane, which can delay the spectral maturation signal even when sucrose is already accumulating. Stress-induced early ripening, sometimes deliberately induced by withholding irrigation or applying chemical ripeners, can produce spectral signatures that resemble natural maturation but at different absolute CCS levels.

Finally, the method requires ground-truth data collected in the same season to calibrate or validate the model. Historical calibration from a different season or a different production region degrades accuracy substantially. Any vendor, including this one, who claims otherwise is overstating what the physics and the published literature support.

Typical figures

Primary spatial resolution	20 m (Sentinel-2 red-edge and SWIR bands); 30 m (Landsat OLI, PRISMA)
Revisit (cloud-free probability)	5-day nominal (Sentinel-2 dual satellite); effective clear-sky revisit in humid tropics typically 10–30 days
Key spectral bands	Red-edge: 705, 740, 783 nm (Sentinel-2 B5/B6/B7); SWIR: 1610, 2190 nm (Sentinel-2 B11/B12); full 400–2500 nm (PRISMA)
Minimum field size for reliable estimate	Approximately 1 ha at 20 m resolution; smaller fields subject to mixed-pixel contamination
Biomass estimation accuracy (published range)	R² 0.70–0.85 against destructive samples in calibrated models; lower without local calibration data
CCS / sucrose estimation accuracy	Weaker than biomass; published correlations typically R² 0.55–0.75; requires seasonal recalibration
Archive depth	Sentinel-2 from 2015; Landsat from 1972 (multispectral) / 2013 (OLI); PRISMA from 2019
Latency from acquisition to product	Sentinel-2 Level-2A available within 3–5 hours of overpass via Copernicus Data Space; processed analytics typically same-day to 48 hours
Coverage	Global; Sentinel-2 covers all agricultural latitudes; PRISMA tasking required for specific sites
Delivery formats	GeoTIFF raster layers, field-polygon summary tables (CSV/GeoJSON), mill scheduling ranked list

Analytics Satellize can run

Seasonal canopy chlorophyll time series	Red-Edge Chlorophyll Index (CIre) and MTCI derived from Sentinel-2 B5/B6/B7 time series, cloud-masked using scene classification layer	Per-field time-series chart and GeoTIFF stack at 20 m, updated on each clear overpass
Stalk biomass estimate map	Empirical regression or PROSAIL inversion relating CIre and LAI to above-ground biomass, calibrated against local destructive samples	GeoTIFF biomass map (t/ha) with per-field polygon summary table, produced at seasonal peak and at harvest window
Harvest-readiness ranking	Phenological curve fitting to CIre time series to identify maturation onset; fields ranked by estimated days to optimal CCS window	Ranked field list (CSV/GeoJSON) with confidence intervals, updated fortnightly during harvest season
SWIR fibre-fraction index	SWIR-ratio index (B11/B7 or equivalent) tracking canopy water depletion and fibre rise; validated against PRISMA hyperspectral snapshots where available	GeoTIFF index layer per overpass; anomaly alert if field diverges from expected maturation trajectory
Within-field maturity variability map	High-resolution (3 m) red-edge index from Planet SuperDove (commercial licence) to identify intra-field maturation heterogeneity	GeoTIFF at 3 m with zones classified as early, mid and late maturity; input to variable-rate harvest scheduling
Multi-year yield trend analysis	Landsat OLI archive (2013–present) used to reconstruct interannual NDVI and SWIR trajectories; anomaly years correlated with ENSO indices and rainfall records	PDF trend report with time-series plots per production zone; GIS layer of yield-anomaly frequency

Sources

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.