Thermal remote sensing measures electromagnetic radiation emitted by Earth's surface in the thermal infrared bands (3-14 um) rather than reflected sunlight. Every object above absolute zero emits thermal radiation governed by its temperature and emissivity (Planck's law). By measuring this emission in atmospheric windows around 3-5 um and 8-14 um, thermal sensors derive land surface temperature (LST) and sea surface temperature (SST). Because thermal emission occurs continuously, thermal sensors operate day and night. Applications span urban heat island mapping, volcanic monitoring, fire detection, and evapotranspiration estimation.
While optical remote sensing measures reflected sunlight, thermal remote sensing measures radiation that the surface itself emits. Reflected energy tells you about surface composition; emitted energy tells you about surface temperature and thermal properties.
The physics is governed by Planck's radiation law: every object above absolute zero emits electromagnetic radiation with a spectral distribution that depends on its temperature and emissivity. Earth's surface, at roughly 288 K, has peak emission near 10 um. Thermal sensors measure this emission in atmospheric windows and convert measured radiance to temperature, provided surface emissivity is known or estimated.
The temperature-emissivity separation problem is the central challenge. Emissivity varies with surface material -- natural surfaces have high emissivity (0.95-0.99), while metals can be much lower. With a single thermal band, you cannot independently determine both temperature and emissivity. Multi-band thermal sensors (like ASTER with 5 thermal bands) use spectral differences to solve for both simultaneously.
Applications exploit the fact that surface temperature responds to physical processes. Urban heat island studies map temperature variations across cities. Fire detection relies on extreme thermal contrast between active fires (800-1200 K) and background (300 K). Sea surface temperature drives ocean circulation models. Evapotranspiration models use LST to estimate water loss -- cooler surfaces are evaporating more. In each case, thermal remote sensing provides information that optical imagery cannot.