Cloud-top temperature (CTT) is the infrared brightness temperature that a satellite assigns to the top of a cloud. It is the coldest reliable estimate the sensor can make of the emitting surface at the cloud's uppermost level.
How the value is produced
An infrared window channel, typically near 10 to 11 micrometers on the GOES-R ABI, measures radiation leaving the top of the cloud. That radiance is inverted through the Planck function to yield brightness temperature, defined in the AMS Glossary as the temperature a hypothetical blackbody would need in order to emit the observed radiation at that wavelength.
The result is a per-pixel temperature field. Colder pixels correspond to emitting surfaces that are colder, which in the troposphere generally means higher altitude. Warmer pixels correspond to lower or thinner cloud tops or to the ground.
How forecasters use it
Cold CTT within a growing cumulonimbus indicates that the top has punched into cold layers of the upper troposphere. A trend of steadily falling CTT is a hallmark of updraft strengthening, and small pockets of CTT dramatically colder than the surrounding anvil are the classic infrared signature of an overshooting top.
For tropical cyclones, cold CTT rings and colder eye walls figure into intensity estimation techniques. For mesoscale convective systems, CTT enhancement helps forecasters ring-fence the coldest, most active portion of the system when severe weather is likely.
Important limits
Brightness temperature is not physical cloud-top temperature in all cases. Semi-transparent cirrus lets radiation from warmer layers below reach the sensor, biasing the reported value warmer than the true top. Thin plumes above a colder anvil can bias the value colder.
CTT is not cloud-top height. Two anvils at the same temperature can sit at different altitudes if their environmental temperature profiles differ. Interpretation is most reliable when CTT is paired with a nearby sounding or a numerical model temperature profile, and when it is watched as a trend on a single storm rather than a single snapshot.