Multispectral Imaging

Multispectral imaging is a technique that captures image data across multiple discrete bands of the electromagnetic spectrum, typically between 3 and 15 bands spanning visible light, near-infrared (NIR), and shortwave infrared (SWIR) wavelengths. Unlike standard RGB photography, which records only the red, green, and blue channels that the human eye can see, multispectral sensors measure reflected and emitted energy in carefully chosen wavelength regions that reveal information invisible to the naked eye — such as vegetation health, soil moisture, and water turbidity.

Multispectral imaging sits between simple RGB capture and hyperspectral imaging on the spectral resolution spectrum. Where hyperspectral sensors record hundreds of very narrow, contiguous bands (often 5–10 nm wide), multispectral instruments use fewer, wider bands (typically 20–180 nm) selected for specific analytical purposes. This trade-off means multispectral data is easier to process, requires less storage, and can be acquired at higher spatial resolutions — making it the practical backbone of most operational Earth observation programs.

A multispectral sensor uses optical filters or detector arrays tuned to specific wavelength ranges. When sunlight illuminates the Earth's surface, different materials — vegetation, water, bare soil, built structures — absorb and reflect energy differently at each wavelength. The sensor records these reflectance values as separate image layers (bands), which can then be combined, compared, and analysed.

By combining bands mathematically, analysts derive spectral indices that quantify surface properties. The most common is NDVI (Normalized Difference Vegetation Index), calculated as (NIR − Red) / (NIR + Red), which exploits the fact that healthy vegetation strongly reflects near-infrared light while absorbing red light for photosynthesis. Other widely used indices include NDWI for water bodies, NDBI for built-up areas, and NDMI for moisture content.

Different band combinations rendered as false-colour composites also reveal distinct features. For example, a NIR–Red–Green composite makes vegetation appear bright red, helping distinguish crop types and forest health, while a SWIR–NIR–Blue combination highlights agricultural fields and soil moisture differences.

Multispectral imaging has inherent trade-offs. Its broad spectral bands can blend distinct absorption features, making it difficult to differentiate materials with subtle spectral differences — for instance, distinguishing specific mineral types or detecting early-stage crop disease before symptoms become pronounced. Hyperspectral imaging is better suited for these tasks. Cloud cover remains a persistent challenge for optical multispectral sensors, as they cannot penetrate clouds, unlike radar systems such as Sentinel-1. Atmospheric effects (haze, aerosols, water vapour) also distort reflectance values and require careful correction. Finally, the spatial resolution of free multispectral data (10–30 m) is insufficient for applications requiring individual-object detection.

The modern era of multispectral Earth observation began on 23 July 1972, when NASA launched the Earth Resources Technology Satellite (ERTS-1, later renamed Landsat 1) carrying the first Multispectral Scanner System (MSS). Designed by Virginia Norwood at Hughes Aircraft Company, the MSS captured four bands at 80 m resolution. Successive Landsat missions refined the technology: Landsat 4 (1982) introduced the Thematic Mapper with 7 bands at 30 m resolution. France's SPOT satellite (1986) added stereo imaging capability. In the 2000s, commercial providers pushed spatial resolution below 1 m. ESA's Sentinel-2 mission (2015–2017) brought the current state of the art for free civilian data: 13 bands, 10 m resolution, and 5-day revisit. Today, constellations like Planet's SuperDove fleet offer daily global multispectral coverage, transforming Earth observation from periodic snapshots into continuous monitoring.