EO·Atlas
Preview build / EO·Atlas v0.9, content still landing
EO·Atlas / Topics / Methodologies / Multispectral optical imaging
Methodology · Multispectral

Multispectral optical imaging

Passive imaging of reflected sunlight in a small number of discrete spectral bands, including single-band panchromatic, from push-broom, whiskbroom and framing cameras. The base optical imaging method behind land, coastal and atmospheric imagery, from very-high-resolution commercial constellations to moderate-resolution weather and land imagers. Band set, ground sampling distance, tasking, orbit and revisit are attributes of an implementation, not separate methods.

How it works

Multispectral optical imaging is used to map land, water and built surfaces from reflected sunlight recorded in a panchromatic band and a small number of discrete spectral bands [1][2]. It depends on daylight, on a clear view of the surface, and on reflective bands in the visible, near-infrared and shortwave-infrared range. Planet states that PlanetScope images only above 10 degrees of sun elevation [3], and the Moderate Resolution Imaging Spectroradiometer (MODIS) cloud mask gives a global cloud fraction of 67 percent [4]. Results also shift with sun and viewing geometry [5]. Demonstrated implementations include Sentinel-2 [1], Landsat 8 and 9 [2], MODIS [6] and PlanetScope [3].

The output is reflectance imagery rather than spectra. Sentinel-2 delivers top-of-atmosphere reflectance (Level-1C) and surface reflectance (Level-2A) products, and the surface product depends on an atmospheric correction step [7]. Thermal infrared bands, such as those on MODIS that extend to 14.385 micrometres, are outside this method, which covers reflective bands only [6].

Spatial detail, swath and revisit vary widely between instruments. WorldView-3 records a 0.31 m panchromatic band and 1.24 m visible and near-infrared bands over a 13.1 km swath at nadir [8]; ESA lists a 30 cm panchromatic band for Pleiades Neo [9]. Sentinel-2 records 13 bands at 10, 20 and 60 m over a 290 km swath, with a 5-day revisit when two satellites operate [1]. Landsat 9 records eight bands at 30 m and one panchromatic band at 15 m, with a 16-day repeat [2]. MODIS pixels range from 250 m to 1 km across a 2330 km swath [6]. The Advanced Baseline Imager on the Geostationary Operational Environmental Satellites R series (GOES-R) scans the full disk every 10 minutes at 0.5 to 2 km [10]. Planet states that PlanetScope covers the landmass near-daily at about 3 m [3]. A panchromatic band can be fused with multispectral bands by pansharpening [11].

Published work using these images includes global forest cover change maps [12], global surface water and its long-term changes [13], the state of the world's beaches [14], the Global Human Settlement Layer produced from Landsat data [15], crop and tree species classification with Sentinel-2 [16], and lithologic mapping with Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data [17].

Cloud is the main constraint: the MODIS cloud fraction is 55 percent over land [4], and ESA states that the Sentinel-2 10:30 overpass time was chosen partly to limit cloud [1]. Masking uses tools such as Fmask for Landsat [18] and the Sen2Cor scene classification, which can confuse bright surfaces and mountain snow with cloud [7]. A changing solar zenith angle can mimic reflectance change, and Landsat views within about 7.5 degrees of nadir still produce small directional effects [5]. Off-nadir tasking degrades resolution; the WorldView-3 panchromatic ground sampling distance rises from 0.31 m to 0.34 m at 20 degrees [8].

Combining sensors needs harmonisation. Sentinel-2 geolocation is 20 m (2 sigma) without ground control points [1], and the Harmonized Landsat and Sentinel-2 data set applies co-registration, bidirectional reflectance distribution function normalisation and bandpass adjustment before joint use [19]. Planet documents four bands on Dove Classic and Dove-R and eight on SuperDove, of which six are interoperable with Sentinel-2 bands [3].

Topics this serves
Topic
Fit
Cloud propertiesfirst choice

Retrieves cloud mask, top height and temperature, phase, optical depth and effective radius from visible, near-infrared and thermal-infrared imager radiances; the basis of every operational cloud product listed.

Disaster damage assessmentfirst choice

VHR optical imagery is the direct observation of damaged buildings and infrastructure.

Coastal erosionsuitable

Waterline position extracted from optical imagery gives shoreline change.

Crop-type mappingsuitable

Multi-date multispectral reflectance is the classification input for 10 m crop maps.

Deforestationsuitable

Multi-band optical time series is the input to 30 m global forest-loss mapping.

Floodingsuitable

Water extent from NIR/SWIR/green bands maps inundation in cloud-free scenes.

Illegal constructionsuitable

Bi-temporal VHR optical imagery shows new structures for comparison against permits and cadastre.

Land cover changesuitable

Supplies the multi-band reflectance time series that land-cover change maps are classified from.

Urban changesuitable

Optical imagery supplies built-up extent at 30 m (Landsat) to sub-metre (VHR).

Crop phenologyadequate

Provides dated reflectance observations; growth stages come from index and fusion methods built on it.

Crop stress and yieldadequate

Red-edge and NIR/SWIR reflectance carry stress signal, but stress retrieval is done by index or fluorescence methods.

Inland wateradequate

Maps surface-water and wetland extent; does not give level, discharge or quality.

Mineral explorationadequate

Broad SWIR bands separate some lithological groups; mineral species need hyperspectral data.

Pasture and grazingadequate

Provides reflectance for grassland condition and mowing detection; biomass comes from index methods.

Wildfireadequate

NIR/SWIR bands map burn scars and severity after the event; active fire needs thermal bands.

Existing implementations
Demonstrated
Capable, undemonstrated
  • AEISS

    Carried from the vegetation-index-mapping edge (capable); multi-band optical imager per live sensor record, record source [wmo-oscar-aeiss].

  • CS-252 Suupistri

    Carried from the vegetation-index-mapping edge (capable); multi-band optical imager per live sensor record, record source [crystalspace-camera-comparison-2024].

  • CS-292 Suupistri-EO

    Carried from the vegetation-index-mapping edge (capable); multi-band optical imager per live sensor record, record source [crystalspace-camera-comparison-2024].

  • Coastal Zone Imager

    Carried from the vegetation-index-mapping edge (capable); multi-band optical imager per live sensor record, record source [pmcncbi-czi-hy1c].

  • MX (Cartosat-3)

    Carried from the vegetation-index-mapping edge (capable); multi-band optical imager per live sensor record, record source [nrsc-cartosat3-brochure].

  • Pléiades Neo Imager

    [esa-pleiades-neo] Pan 30 cm, six MS bands incl. deep blue and red edge at 1.2 m, 14 km swath, two satellites.

  • PAIS Optical Payloadpending review

    Carried from the vegetation-index-mapping edge (capable-pending-review); multi-band optical imager per live sensor record, record source [remondo-technology].

  • Valk VHRpending review

    Carried from the vegetation-index-mapping edge (capable-pending-review); multi-band optical imager per live sensor record, record source [cosine-valk-ministerial].

  • CO3D Optical Imager

    [record:co3d-cnes] 50 cm VNIR imager on CO3D (launched 2025); no delivered-imagery source checked, so capable not demonstrated.

  • DragonEye Imager

    [record:eoportal-eossat1] 1.4 m pan and 2.8 m MS imager; eoPortal EOS SAT-1 page cited by the record, but no mission_payloads link exists, so capable not demonstrated.

  • MultiScape50

    [record:multiscape50-page] Operator datasheet: 1 pan + 7 VNIR band pushbroom, 30 m GSD; no flight evidence checked.

  • MultiScape100

    [record:multiscape100-page] Operator datasheet: 1 pan + 7 VNIR band pushbroom, 4.75 m GSD; no flight evidence checked.

  • MultiScape200

    [record:multiscape200-page] Operator datasheet: 8-channel MS pushbroom, 1.5 m GSD; no flight evidence checked.

  • TriScape50

    [record:triscape50-page] Operator datasheet: RGB video imager (framing), 30 m GSD.

  • TriScape100

    [record:triscape100-page] Operator datasheet: RGB video imager (framing), 4.75 m GSD.

  • TriScape200

    [record:triscape200-page] Operator datasheet: RGB+NIR snapshot and video imager, 1.5 m GSD.

Related methodologies
  • Vegetation index mapping

    Derives vegetation health, density, and phenology indicators (NDVI, EVI, SAVI) from red and NIR band ratios; good for crop monitoring, deforestation alerts, and land cover change detection.

  • Ocean colour / water-leaving radiance

    Retrieves chlorophyll-a concentration, CDOM, and suspended sediment from spectrally resolved water-leaving radiance; good for marine primary production, harmful algal bloom detection, and coastal water quality.

  • Optical sun-glint surface imaging

    Images the sea surface in sun-glint geometry, where surface-slope modulation of reflected sunlight reveals wave patterns and roughness; supports wave-spectrum and roughness estimation under clear-sky, suitable-illumination conditions. A complementary optical method, not an all-weather baseline.

  • Nightlights radiometry

    Measures calibrated night-time visible to near-infrared radiance from low-light imagers, supporting economic activity, electrification, outage, urbanisation, and light-pollution analyses.

Sources
Methodology

Edited from public sources. Last reviewed 2026-10-21 by SpectraWorks editorial. See the data dictionary for field definitions.

Cite or share
https://eo-atlas.org/topics/methodologies/multispectral-optical-imaging
Markdown twin →
How to cite this page

Plain text

Multispectral optical imaging. EO-Atlas, SpectraWorks B.V., last reviewed 2026-10-21. https://eo-atlas.org/topics/methodologies/multispectral-optical-imaging (CC BY 4.0)

BibTeX

@misc{eoatlas:topics:methodologies:multispectral-optical-imaging,
  title = {Multispectral optical imaging},
  author = {{EO-Atlas}},
  publisher = {SpectraWorks B.V.},
  year = {2026},
  note = {Last reviewed 2026-10-21. CC BY 4.0},
  url = {https://eo-atlas.org/topics/methodologies/multispectral-optical-imaging}
}