← Field Manual

SAT-002

Sentinel-2

ESA's multispectral workhorse: 13 bands, 10-meter resolution, and the backbone of global land monitoring

Plain Summary

Sentinel-2 is a pair of optical satellites operated by the European Space Agency (ESA) as part of the Copernicus Earth observation program. Each satellite carries a Multispectral Instrument (MSI) that captures images in 13 spectral bands, from visible light through near-infrared to shortwave infrared, at resolutions of 10, 20, and 60 meters. With two satellites in orbit (Sentinel-2A and 2B), the combined revisit time is approximately 5 days at the equator and 2-3 days at mid-latitudes. The data is free and open access. Sentinel-2 is the most widely used optical satellite for land monitoring, agriculture, forestry, water resources, and disaster response worldwide.


Why It Matters

Before Sentinel-2, the options for free, high-resolution, multispectral satellite imagery were limited. Landsat provided 30-meter resolution with a 16-day revisit, excellent for long-term studies but too coarse and too infrequent for many operational applications. Commercial satellites offered higher resolution but at costs that excluded most researchers, governments of developing nations, and non-governmental organizations.

Sentinel-2 changed the economics and temporal density of Earth observation simultaneously. Ten-meter resolution in its visible and near-infrared bands is sufficient to monitor individual agricultural fields, detect small-scale deforestation, map urban expansion, and identify water bodies. The 5-day revisit means that even with cloud cover, usable imagery is available for most locations every 1-2 weeks, often enough for operational monitoring rather than just periodic assessment.

The combination of spatial resolution, spectral range, temporal frequency, and open data access made Sentinel-2 the default satellite for a vast range of applications. More scientific papers cite Sentinel-2 data than any other single Earth observation mission.


The Multispectral Instrument (MSI)

The MSI captures reflected sunlight in 13 spectral bands, each designed to measure specific properties of the Earth's surface and atmosphere:

10-Meter Bands

Band Name Wavelength (nm) Primary Use
B2 Blue 490 Atmospheric scattering, water bodies, coastal studies
B3 Green 560 Vegetation vigor, peak green reflectance
B4 Red 665 Chlorophyll absorption; the key band for vegetation stress
B8 NIR (broad) 842 Vegetation structure, biomass, water boundary detection

These four bands form the core of most Sentinel-2 analyses. The B4/B8 ratio is the basis of NDVI, the most widely used vegetation index. The 10-meter resolution at these bands makes Sentinel-2 competitive with commercial satellites for many agricultural and environmental applications.

20-Meter Bands

Band Name Wavelength (nm) Primary Use
B5 Red Edge 1 705 Vegetation red edge; sensitive to chlorophyll and nitrogen
B6 Red Edge 2 740 Canopy structure, LAI estimation
B7 Red Edge 3 783 Leaf area, canopy water content
B8A NIR (narrow) 865 Vegetation / water feature discrimination (narrower than B8)
B11 SWIR 1 1610 Moisture content, burn severity, snow/cloud discrimination
B12 SWIR 2 2190 Geological mapping, soil and vegetation moisture, burn severity

The three red edge bands (B5, B6, B7) are unique to Sentinel-2 among free missions. The red edge, the sharp transition in vegetation reflectance between red absorption and NIR reflection, is highly sensitive to chlorophyll content, nitrogen status, and vegetation health. These bands enable more precise vegetation monitoring than is possible with Landsat's broader red and NIR bands.

The SWIR bands (B11, B12) are essential for distinguishing clouds from snow, mapping burn severity (used in the Normalized Burn Ratio), and detecting moisture stress in vegetation and soil.

60-Meter Bands

Band Name Wavelength (nm) Primary Use
B1 Coastal Aerosol 443 Atmospheric correction, aerosol retrieval
B9 Water Vapour 945 Atmospheric water vapor estimation
B10 Cirrus 1375 Cirrus cloud detection

These bands are not intended for surface analysis. They support atmospheric correction and cloud masking, the preprocessing that converts raw top-of-atmosphere measurements to analysis-ready surface reflectance.


Spectral Indices

Sentinel-2's band configuration enables computation of dozens of standardized spectral indices, each designed to highlight specific surface properties:

NDVI (Normalized Difference Vegetation Index) = (B8 − B4) / (B8 + B4) The most widely used vegetation index. Values range from -1 to 1, with healthy vegetation typically 0.3-0.8. Based on the principle that chlorophyll absorbs red light while leaf structure reflects NIR.

NDWI (Normalized Difference Water Index) = (B3 − B8) / (B3 + B8) Highlights water bodies. Water absorbs NIR strongly, producing high NDWI values. Used for water extent mapping and flood detection.

NDMI (Normalized Difference Moisture Index) = (B8 − B11) / (B8 + B11) Sensitive to vegetation water content through SWIR absorption. Used for drought monitoring and irrigation assessment.

NBR (Normalized Burn Ratio) = (B8 − B12) / (B8 + B12) Distinguishes burned from unburned areas. The difference in NBR between pre-fire and post-fire imagery (dNBR) is the standard metric for burn severity mapping.

NDSI (Normalized Difference Snow Index) = (B3 − B11) / (B3 + B11) Exploits snow's high visible reflectance and strong SWIR absorption to map snow cover.

Red Edge NDVI variants using bands B5, B6, or B7 instead of B4 provide increased sensitivity to chlorophyll content and are used in precision agriculture for crop health assessment and nitrogen status estimation.


Data Products and Access

Sentinel-2 data is distributed at two primary processing levels:

Level-1C — Top-of-atmosphere reflectance. Radiometrically and geometrically corrected, including orthorectification using a digital elevation model. This is what the sensor "sees" before atmospheric correction.

Level-2A — Surface reflectance. Atmospheric correction applied using the Sen2Cor processor, producing bottom-of-atmosphere reflectance with a Scene Classification Layer (SCL) that classifies each pixel as: vegetation, not-vegetated, water, cloud high/medium/low probability, cloud shadow, thin cirrus, snow/ice, saturated/defective, or dark area.

Level-2A is the standard analysis-ready data product for most applications.

Data is tiled using the Military Grid Reference System (MGRS) in 100×100 km granules. Coverage extends from 56°S to 84°N latitude.

Access points:

  • Copernicus Data Space Ecosystem (dataspace.copernicus.eu) — ESA's primary portal
  • Copernicus Browser — Visual search and download
  • STAC-compatible APIs for programmatic access
  • Commercial cloud providers (AWS, Google, Microsoft) host complete Sentinel-2 archives

Cross-Sensor Harmonization

Sentinel-2 is frequently used in combination with Landsat 8/9 to increase temporal density. However, the two missions differ in ways that complicate direct comparison:

Spectral response functions differ. Sentinel-2's Band 4 (Red, 665nm center, 30nm width) and Landsat 8's Band 4 (Red, 655nm center, 37nm width) measure overlapping but not identical portions of the spectrum. These differences produce systematic offsets in derived indices; NDVI computed from Sentinel-2 will differ from NDVI computed from Landsat for the same surface.

Spatial resolution differs. Sentinel-2's 10m pixels cover 100 m² each, while Landsat's 30m pixels cover 900 m². Combining the two requires resampling, which introduces scale-dependent effects.

View angle differs. Sentinel-2 has a wider swath (290 km vs 185 km) with greater off-nadir viewing at swath edges, introducing angular effects in reflectance (BRDF).

The Harmonized Landsat Sentinel-2 (HLS) dataset, produced by NASA, addresses some of these issues by adjusting spectral response functions, resampling to a common grid, and applying BRDF correction. Fabric takes this further by incorporating provenance tracking and extending harmonization to additional sensor types including SAR.


Limitations

Sentinel-2 is a passive optical sensor, which imposes fundamental constraints:

Cloud cover blocks observation. No optical sensor can see through clouds. Persistently cloudy regions (tropical forests, monsoon zones) may have only a handful of clear acquisitions per year. SAR is the primary complement for cloud-affected monitoring.

Nighttime observation is not possible. MSI measures reflected sunlight, so acquisitions occur only during daytime passes. This limits fire detection (nighttime fires are invisible to Sentinel-2 but detectable by thermal sensors on other missions like VIIRS).

Spatial resolution of 10-20m is insufficient for some applications. Building-level analysis, individual tree detection, and infrastructure inspection typically require sub-meter commercial imagery.

Temporal resolution, while dramatically better than Landsat, is still insufficient for monitoring highly dynamic phenomena. A 5-day revisit means fast-moving floods may peak between passes. Time-series analysis at sub-weekly cadence requires combining with other optical missions or SAR.

Atmospheric correction uncertainty affects all derived products. The Sen2Cor processor relies on models of atmospheric state that are imperfect, particularly over bright surfaces (deserts, snow) and in complex topography. See Atmospheric Correction.


The Fabric Connection

Sentinel-2 is the primary optical data source in Fabric's processing pipeline. Fabric's fire detection, vegetation monitoring, and water extent patterns all begin with Sentinel-2 Level-1C or Level-2A data.

Fabric adds value beyond standard Sentinel-2 processing in three ways. First, it handles the full ARD pipeline (atmospheric correction, cloud masking, geometric alignment, and quality assessment) with provenance tracking at every step. Second, it harmonizes Sentinel-2 with other sensors (Landsat, SAR, thermal) to produce composite observations that no single sensor can achieve alone. Third, through Observational Grammar, it treats every Sentinel-2 measurement as a claim with explicit uncertainty rather than as ground truth — the atmospheric correction is documented as an interpretation, not a fact.


Satellites & Platforms: Sentinel-1 · Landsat Program · MODIS / VIIRS · Open Data Programs · Orbit Types

Sensors & Physics: Multispectral Imaging · Atmospheric Correction · Radiometric Principles · Spatial/Spectral/Temporal Resolution

Earth Observation: Vegetation Health (NDVI/EVI) · Fire Detection & Monitoring · Flood & Water Extent · Deforestation & Land Cover Change

Data & Architecture: Analysis-Ready Data · Harmonization · STAC · Cloud-Optimized GeoTIFF · Coordinate Reference Systems


References

[1] Drusch, M., et al. (2012). "Sentinel-2: ESA's Optical High-Resolution Mission for GMES Operational Services." Remote Sensing of Environment, 120, 25-36. doi:10.1016/j.rse.2011.11.026

[2] ESA (2024). "Sentinel-2 User Handbook." European Space Agency. sentinel.esa.int

[3] Main-Knorn, M., et al. (2017). "Sen2Cor for Sentinel-2." Proc. SPIE 10427, Image and Signal Processing for Remote Sensing XXIII.

[4] Claverie, M., et al. (2018). "The Harmonized Landsat and Sentinel-2 Surface Reflectance Data Set." Remote Sensing of Environment, 219, 145-161.

[5] Phiri, D., et al. (2020). "Sentinel-2 Data for Land Cover/Use Mapping: A Review." Remote Sensing, 12(14), 2291. doi:10.3390/rs12142291


Further Reading

Sentinel-2 User Handbook — ESA, 2024 The official reference. Includes detailed spectral response functions, radiometric performance, orbit parameters, and data product descriptions.

Remote Sensing of Environment, Volume 120 — Special Issue on Sentinel-2, 2012 The original mission description papers covering instrument design, calibration strategy, and planned applications.

Harmonized Landsat Sentinel-2 (HLS) User Guide — NASA, 2023 How NASA produces harmonized data products from the two missions. Essential reading for understanding the cross-sensor challenges Fabric addresses.


Entry SAT-002 · Created February 2026 · Contributors: M33 Team · License: CC BY-SA 4.0