Multispectral imagers
Multispectral imagers split reflected sunlight into 4–13 discrete bands, letting analysts distinguish crop stress from bare soil, turbid water from healthy reef. They are the workhorse payload for national monitoring programmes where spectral identity matters more than centimetre-scale detail.
What a band actually buys you
A panchromatic sensor integrates all visible wavelengths into a single brightness value. A multispectral imager splits that same incoming light into narrow slices, typically 20–100 nm wide, each recorded as a separate image plane. The ratio or combination of those planes is where the information lives. The normalised difference vegetation index (NDVI), for instance, is simply the ratio of near-infrared reflectance to red reflectance; healthy chlorophyll absorbs red and strongly reflects NIR, so the ratio rises sharply. No amount of spatial resolution in a single panchromatic band can produce that number.
The practical consequence for a government buyer is that band count and band placement determine which applications the satellite can actually serve. Four bands (blue, green, red, NIR) cover most vegetation and land-cover work. Adding a red-edge band around 705–740 nm improves early detection of crop stress before visible yellowing appears. A shortwave infrared band at 1.6 µm or 2.2 µm separates moisture content, mineral composition and cloud from ice. Each additional band adds optical complexity, focal-plane area, data volume and calibration burden. The mission architecture should start with the application list, not the band catalogue.
The aperture decides the price
Ground sampling distance is set by the ratio of focal length to detector pixel pitch, scaled by orbital altitude. Sentinel-2A and 2B, flying at 786 km, achieve 10 m GSD in their four main bands using a three-mirror anastigmat telescope with a 60 cm aperture and 1.5 µm pixel pitch detectors. That aperture, at that altitude, in a 290 km swath, demands a satellite in the 1,100 kg class. A 6U or 16U smallsat at 500 km altitude with a 5 cm aperture will deliver roughly 20–30 m GSD in comparable bands, a swath of perhaps 20–40 km, and a radiometric signal-to-noise ratio that is meaningfully lower.
The trade is not simply resolution versus cost. Swath width drives revisit: Sentinel-2's 290 km swath gives a five-day repeat at the equator with two satellites. A smallsat constellation of 20 nodes with 30 km swaths can match or beat that revisit but requires a more complex ground processing pipeline to mosaic the strips. For a single-satellite sovereign programme, a wide swath at moderate resolution is almost always more useful than a narrow swath at fine resolution, because consistent, cloud-free coverage of a national territory is harder to achieve than most buyers expect.
Radiometric quality is the part that gets undersold
Spatial resolution appears in marketing materials. Radiometric quality rarely does. It should. A pixel value in a multispectral image is only analytically useful if it can be converted to a physically meaningful surface reflectance, and that conversion requires careful attention to detector uniformity, stray light suppression, absolute calibration against known targets, and atmospheric correction. Sentinel-2 carries on-board diffuser panels for vicarious calibration and achieves absolute radiometric uncertainty below 3% across most bands, a figure that took years of post-launch characterisation to validate.
A smallsat multispectral payload will typically achieve 5–10% absolute radiometric uncertainty without dedicated calibration infrastructure, which is acceptable for change detection within a single sensor's time series but problematic if you need to compare values across sensors or against historical archives. Buyers planning to build long time series for agriculture or water quality monitoring should require the supplier to specify signal-to-noise ratio per band at a defined reference radiance, not just GSD. An SNR below roughly 100:1 in the NIR will produce noisy NDVI values over low-reflectance targets such as water or dark soils.
Where multispectral genuinely struggles
Cloud cover is the dominant operational limit. Passive optical sensors, multispectral included, record reflected sunlight. A 1 km thick cumulus cloud produces a reflectance of roughly 0.7–0.9 across visible and NIR bands, completely saturating the signal from the surface below. In persistently cloudy regions, a single satellite may return fewer than ten cloud-free acquisitions per year over a given location. Constellation size and orbit selection can improve the odds but cannot eliminate the physics.
Spectral resolution is the second honest limit. Thirteen bands is not the same as hundreds of contiguous bands. A multispectral imager cannot distinguish between two mineral species with similar broadband reflectance profiles, cannot reliably separate phytoplankton species in coastal water, and cannot detect narrow absorption features associated with specific pollutants. Those tasks belong to hyperspectral instruments. Multispectral is the right choice when the target phenomenon produces a broad spectral contrast across the bands available; it is the wrong choice when the phenomenon is spectrally subtle.
Finally, atmospheric correction over complex terrain or coastal water remains an unsolved operational problem for automated pipelines. Sun-angle effects, adjacency effects from bright surfaces and aerosol variability all introduce errors that are hard to remove without ground truth. Buyers should treat surface reflectance products as estimates with uncertainty bands, not as measurements.
Band selection for a national programme
Most national monitoring mandates collapse into four application clusters: agriculture and food security, land cover and urban change, water quality and coastal zone, and disaster response. A six-band configuration covering blue (450–520 nm), green (530–590 nm), red (630–690 nm), red-edge (700–740 nm), NIR (760–900 nm) and SWIR-1 (1550–1700 nm) addresses all four clusters adequately. Adding a second SWIR band at 2.0–2.3 µm improves mineral and soil moisture discrimination but adds cost and cooling complexity. A coastal/aerosol band below 450 nm improves atmospheric correction over water but is technically demanding to calibrate.
The Satellize crop-estimation programme for the Kingdom of Tonga demonstrated that a carefully chosen four-band configuration, combined with rigorous field calibration, can produce actionable yield estimates from smallsat-class imagery. The limiting factor was not band count but calibration discipline and the availability of ground-truth plots. That experience informs our standard recommendation: specify the minimum band set that serves your application list, invest the savings in calibration infrastructure and ground-truth collection, and plan for the processing pipeline before the satellite is built.
Engineering parameters
| Typical band count | 4–13 bands (visible, NIR, SWIR; exact selection mission-dependent) |
| Ground sampling distance (smallsat class, ~500 km alt) | 15–30 m typical; sub-10 m achievable with larger aperture |
| Ground sampling distance (Sentinel-2 class, ~786 km alt) | 10 m (main bands), 20 m (red-edge, SWIR), 60 m (coastal/aerosol) |
| Swath width | 20–40 km (smallsat class) to 290 km (Sentinel-2 class) |
| Payload mass (smallsat class) | 1–8 kg depending on aperture and band count |
| Payload power (smallsat class) | 5–30 W during imaging |
| Absolute radiometric uncertainty | 3–5% (well-calibrated large platform); 5–12% (smallsat without dedicated calibration) |
| Signal-to-noise ratio (NIR at reference radiance) | 100:1 minimum acceptable for vegetation indices; >200:1 preferred |
| Raw data rate (6-band, 10 m GSD, 40 km swath) | 200–600 Mbit/s before compression; typically 4:1 lossless compression applied on-board |
| Detector technology | Silicon CMOS or CCD (VIS/NIR); InGaAs array (SWIR bands) |
One contract, one accountable engineer
Commissioned as one programme, not a stack of contracts: spacecraft, launch, ground segment, mission control, training and handover are priced together. Source-access terms and audit rights are agreed in writing before signature. Request a band-selection review for your mission.