Thermal infrared imagers
Thermal infrared payloads detect emitted heat rather than reflected sunlight, making them indispensable for night imaging, fire detection and industrial monitoring. Resolution is fundamentally constrained by wavelength, and that trade-off defines every mission choice.
Emitted heat, not reflected light
Visible and multispectral cameras are passive instruments that depend on the Sun. A thermal infrared imager is also passive, but its energy source is the target itself. Every object above absolute zero emits electromagnetic radiation. At terrestrial temperatures, that emission peaks in the longwave infrared band, roughly 8 to 14 micrometres (LWIR), with a secondary window in the midwave band at 3 to 5 micrometres (MWIR). A thermal payload images that self-emission directly, which is why it works at night, through smoke, and in conditions where a conventional optical sensor sees nothing useful.
The physical principle is Planck's law. A surface at 300 K emits peak radiation near 10 micrometres. A steel plant running at 900 K shifts its peak toward the MWIR. Fire fronts, gas flares and volcanic vents are bright in MWIR precisely because their temperatures are high enough to push significant emission into that band. LWIR is the workhorse for ambient-temperature targets: urban heat islands, agricultural stress, sea-surface temperature, building energy loss. The two bands are complementary, and some payloads carry both.
The aperture decides the price
Resolution in any imaging system is set by the ratio of wavelength to aperture diameter. LWIR wavelengths are roughly twenty times longer than visible light. To match the ground resolution of a 50 cm panchromatic camera, a thermal imager would need an aperture twenty times larger, which is impractical on any small platform. In practice, spaceborne thermal imagers deliver ground sample distances of 60 to 100 metres from low Earth orbit for systems in the Landsat class, and 30 to 50 metres for larger dedicated instruments. Landsat 8's Thermal Infrared Sensor (TIRS) achieves 100 m GSD from a 705 km orbit using a 10.6 cm aperture and a 640 × 512 pixel quantum-well infrared photodetector array cooled to approximately 43 K. That cryogenic cooling is what enables the sensitivity needed for subtle land-surface temperature differences.
Smallsat microbolometer payloads operate without cooling, which dramatically reduces mass, power and cost. Uncooled microbolometers typically achieve noise-equivalent temperature differences (NETD) of 50 to 100 mK, compared with 20 to 30 mK for cooled systems. The trade is sensitivity and, consequently, the ability to detect small temperature anomalies. A 12-metre GSD uncooled thermal smallsat is achievable, but detecting a 0.5 K sea-surface temperature anomaly is not. For fire detection, where temperature contrasts are tens or hundreds of kelvin, an uncooled microbolometer is perfectly adequate. For agricultural stress mapping, where the signal of interest may be 1 to 2 K, a cooled detector is usually necessary.
What thermal data actually tells an operator
Land-surface temperature derived from LWIR is used to map urban heat islands, monitor reservoir evaporation, detect geothermal anomalies and estimate crop water stress through the relationship between canopy temperature and evapotranspiration. Sea-surface temperature from LWIR is a direct input to fisheries management, hurricane intensity forecasting and ocean current monitoring. Both applications require absolute temperature accuracy of roughly 0.5 to 1 K, which demands careful atmospheric correction using simultaneous measurements of water vapour and aerosol loading.
Fire detection is the most operationally time-sensitive application. MWIR is preferred because active combustion at 600 to 1000 K radiates intensely in that band, allowing sub-pixel fire detection: a fire occupying a fraction of one pixel still saturates the detector response and can be flagged algorithmically. The European Space Agency's Sentinel-3 SLSTR instrument uses a dedicated 1 km MWIR fire channel for exactly this reason. Industrial applications include gas-flare monitoring for emissions accounting, pipeline leak detection (where escaping gas cools the surface), and steel and cement plant thermal auditing.
Cloud, atmosphere and the limits of thermal sensing
Thermal infrared does not penetrate cloud. This is the single largest operational constraint. Cloud cover is not just a nuisance; it is optically opaque in the LWIR band, and the instrument images cloud-top temperature rather than the surface below. For persistent cloud regions, such as the humid tropics, thermal revisit is effectively lower than orbital mechanics suggest. A nominal daily revisit becomes three to five usable passes per month in practice. Mission planners for agricultural or fire applications in cloudy regions must either accept data gaps, build constellations with higher revisit to improve statistical coverage, or fuse thermal data with SAR to fill temporal gaps.
Atmospheric correction is non-trivial. Water vapour absorbs strongly in parts of the LWIR window, and without accurate atmospheric profiles, retrieved surface temperatures carry errors of 2 to 5 K. Operational systems such as MODIS and SLSTR use split-window algorithms that exploit the differential absorption between two adjacent LWIR channels to estimate and remove atmospheric effects. A single-channel thermal payload cannot apply split-window correction and is therefore limited in absolute temperature accuracy. Stray light and self-emission from the instrument itself also contaminate the signal; Landsat TIRS suffered a known stray-light artefact in its first version that required a correction algorithm and a redesigned baffle on TIRS-2. These are not hypothetical risks. They are documented problems on well-funded programmes.
Choosing between cooled and uncooled for a sovereign programme
The decision tree is straightforward once mission requirements are honest. Cooled detectors (quantum-well infrared photodetectors, mercury cadmium telluride arrays) deliver the sensitivity needed for sea-surface temperature, agricultural stress and subtle geothermal mapping. They add mass (typically 15 to 40 kg for the cryocooler and detector assembly), power (30 to 80 W), cost and a mechanical failure mode: the Stirling-cycle cryocooler has a finite service life, usually quoted at 5 to 7 years of continuous operation. For a 10-year mission, cryocooler lifetime is a design constraint, not an afterthought.
Uncooled microbolometer arrays have matured considerably. Commercial units derived from defence production lines are available in formats from 640 × 512 to 1280 × 1024 pixels, with pixel pitches of 12 to 17 micrometres, at masses under 500 g for the detector alone. Integrated into a smallsat with a 150 to 200 mm aperture, they can achieve 30 to 50 m GSD from a 500 km orbit with NETD around 80 mK. For a government wanting fire monitoring, industrial surveillance or night-time activity detection, an uncooled constellation of three to six smallsats can provide sub-daily revisit at a fraction of the cost of a single large cooled instrument. The honest caveat: if the science requirement demands 0.3 K absolute accuracy, uncooled is not the right answer.
Engineering parameters
| Spectral bands | LWIR: 8–14 µm; MWIR: 3–5 µm; some payloads carry both |
| Ground sample distance (GSD) | 30–100 m for Landsat/Sentinel class; 30–50 m achievable on smallsat with 150–200 mm aperture from 500 km |
| NETD (cooled detector) | 20–30 mK typical (e.g. Landsat TIRS QWIP array) |
| NETD (uncooled microbolometer) | 50–100 mK typical for commercial smallsat-grade units |
| Payload mass (cooled system) | 15–60 kg including cryocooler; Landsat TIRS-2 approximately 29 kg |
| Payload mass (uncooled smallsat) | 1–5 kg integrated detector and optics |
| Power (cooled system) | 30–80 W; cryocooler dominates |
| Power (uncooled smallsat) | 3–15 W |
| Cryocooler service life | 5–7 years continuous operation (Stirling cycle); mission-life constraint for 10-year programmes |
| Absolute temperature accuracy | 0.5–1 K with split-window correction (dual-channel); 2–5 K error possible with single-channel and no atmospheric profile |
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 thermal mission trade study.