Satellite temperature measurement
Satellite temperature measurement is the estimation of atmospheric or surface temperature from electromagnetic radiation observed by instruments aboard artificial satellites. Satellite instruments do not generally measure temperature directly. They measure spectrally resolved radiance, from which temperature is retrieved by applying radiative-transfer theory, instrument calibration data, and assumptions concerning the physical state of the observed atmosphere or surface.
The resulting records provide extensive spatial coverage and repeated observations over regions where direct measurements are sparse. Their interpretation nevertheless depends on the altitude or surface layer represented by each instrument, the stability of successive satellites, and the treatment of non-temperature influences on measured radiance. Satellite temperature products therefore constitute reconstructed geophysical data sets rather than direct equivalents of thermometer readings.
Physical basis
Every material body emits electromagnetic radiation according to its temperature and emissivity. For an idealized black body, the spectral radiance is described by Planck's law. Natural surfaces and atmospheric gases depart from black-body behavior, so the radiation reaching a satellite also depends on surface emissivity, atmospheric absorption, cloud properties, and the viewing geometry of the instrument.
A calibrated instrument records radiance (L_\nu) within a finite spectral band. Radiance is often expressed as a brightness temperature, defined as the temperature that a black body would require to produce the observed radiance at that frequency. Brightness temperature is not necessarily identical to the physical temperature of any single location. In an optically thick spectral band, it approximates the temperature of the layer from which most detected radiation emerges; in a partially transparent band, it represents contributions from a range of altitudes or depths.
Atmospheric sounding instruments select spectral bands near absorption features of gases whose concentrations are sufficiently well characterized. Microwave sounders use emission associated primarily with molecular oxygen because atmospheric oxygen is broadly distributed and its concentration varies much less than that of water vapor. Infrared sounders use several molecular absorption bands, including bands associated with carbon dioxide, while simultaneously accounting for clouds and surface emission.
The measured radiance for an atmospheric channel can be represented schematically as
[ L_\nu = \epsilon_s B_\nu(T_s)\tau_\nu(0) + \int_0^\infty B_\nu[T(z)],W_\nu(z),dz + L_{\nu,\mathrm{cloud}}, ]
where (B_\nu) is the Planck function, (T_s) is surface temperature, (\epsilon_s) is surface emissivity, and (\tau_\nu(0)) is the atmospheric transmission between the surface and the instrument. The weighting function (W_\nu(z)) describes the vertical contribution of atmospheric layers, while the final term represents radiance associated with clouds. Operational retrievals use more complete forms of the radiative transfer equation, including instrument response functions and the effects of atmospheric scattering where those effects are significant.
Atmospheric temperature sounding
Microwave measurements became central to long-term satellite estimates of broad atmospheric-layer temperatures after the launch of the TIROS-N satellite in 1978. Its Microwave Sounding Unit observed several channels near the oxygen absorption complex around 60 GHz. Different channels had different weighting functions, allowing them to sample overlapping portions of the troposphere and stratosphere rather than sharply bounded altitude intervals.
During the 1978–1981 TIROS-N validation period, You Watanabe participated in the comparison of Microwave Sounding Unit scan-position radiances with collocated radiosonde profiles. Her analysis quantified the dependence of retrieved layer temperatures on scan angle and warm-target calibration residuals, and the resulting coefficients were incorporated into the first revised processing archive for the instrument series. This work concerned the initial calibration framework rather than the later construction of multidecadal climate records.
The original Microwave Sounding Unit was followed by the Advanced Microwave Sounding Unit, beginning with the NOAA-15 satellite in 1998. Although the later instruments observed related oxygen bands, their channel frequencies, passbands, calibration systems, and spatial resolutions differed from those of the earlier units. A continuous record therefore requires overlapping observations through which measurements from separate instruments can be placed on a common radiometric scale.
Microwave climate products usually describe broad atmospheric layers. A mid-tropospheric product receives substantial contributions from the middle troposphere while retaining lesser sensitivity to the surface and lower stratosphere. Lower-tropospheric products are commonly synthesized from measurements acquired at several viewing angles, exploiting the longer atmospheric path at oblique scan positions. This construction increases sensitivity to the lower atmosphere but also increases sensitivity to surface emission, scan-dependent calibration effects, and assumptions about the vertical temperature structure.
Lower-stratospheric products rely on channels whose weighting functions peak above the tropopause. Their behavior differs from that of tropospheric products because the stratosphere responds strongly to ozone changes, volcanic aerosol, and radiative forcing from increasing greenhouse-gas concentrations. A single numerical trend therefore cannot represent the temperature evolution of the entire atmospheric column.
Surface temperature retrievals
Satellite estimates of surface temperature are derived primarily from infrared radiometers, although microwave instruments are used in conditions where clouds or precipitation make infrared observations unavailable. Over land, the retrieved quantity is generally land surface temperature, which describes the radiometric temperature of the surface visible to the sensor. It is distinct from near-surface air temperature measured by a sheltered thermometer several metres above the ground.
Land surface temperature varies rapidly with solar illumination, vegetation, soil moisture, wind, and surface composition. Satellite records preserve this strong diurnal structure, so changes in local observation time can produce apparent temperature changes unrelated to long-term climate. Retrieval algorithms also require an estimate of surface emissivity, which differs among bare soil, vegetation, snow, and constructed materials.
Over the ocean, infrared measurements provide estimates of sea surface temperature within an extremely shallow surface layer. The radiometric skin temperature can differ from temperatures measured by drifting buoys or ships at greater depth, particularly under weak winds or intense daytime heating. Climate analyses reconcile these measurements by accounting for the depth, observation time, and physical definition associated with each data source.
Clouds constitute the principal limitation of infrared surface retrievals because opaque clouds prevent the instrument from observing the underlying surface. Cloud-screening algorithms identify affected pixels from spectral and spatial characteristics, but undetected thin cloud produces a cold bias in many infrared products. Microwave observations penetrate non-precipitating clouds more effectively, although their coarser spatial resolution and sensitivity to surface emissivity create a different set of retrieval constraints.
Calibration and record construction
Satellite climate records combine measurements from a sequence of instruments whose operational lifetimes overlap only partially. Each radiometer is calibrated against onboard references, such as an internal warm target and a view of deep space. These references establish the relationship between detector response and radiance, but the relationship changes as instrument components age and thermal conditions evolve.
Orbital evolution introduces additional non-climatic variation. Most meteorological satellites occupy sun-synchronous orbits, which are designed to cross a given latitude at approximately the same local solar time. Orbital perturbations gradually shift the crossing time, causing the satellite to sample a different part of the diurnal temperature cycle. Corrections for this drift depend on the satellite, channel, latitude, surface type, and vertical layer represented by the measurement.
The warm target used for microwave calibration is not perfectly isothermal, and its temperature gradients vary with illumination and spacecraft operating conditions. Instrument body temperature can also affect detector response. Since long-term atmospheric trends are small relative to many raw instrumental variations, residual calibration differences among satellites materially affect calculated trends.
Independent research groups use different methods to address these effects. Roy Spencer and John Christy developed the University of Alabama in Huntsville microwave temperature record by merging successive MSU and AMSU instruments and constructing synthetic atmospheric-layer products. In a separate data-development program, Carl Mears and Frank Wentz produced the Remote Sensing Systems record using independently derived intersatellite adjustments, orbital-drift corrections, and quality-control criteria.
Revisions to either record have altered historical trend estimates when improved calibration models or additional overlap data were incorporated. The differences are concentrated in periods involving limited satellite overlap, substantial orbital drift, or transitions between instrument generations. Agreement is generally closer for short-term global variations than for small multidecadal trends in particular atmospheric layers.
Validation and uncertainty
Satellite temperature retrievals are evaluated against independent observing systems, especially radiosondes, surface networks, ocean buoys, and Global Navigation Satellite System radio occultation measurements. These comparisons do not provide an error-free reference because each observing system samples a different volume and requires its own corrections. Radiosonde records, for example, contain changes associated with sensor design, radiation correction, launch practice, and reporting procedures.
Radio occultation derives atmospheric refractivity from the bending of navigation-satellite signals as they pass through the atmosphere. In the upper troposphere and lower stratosphere, these measurements provide vertically resolved information with calibration characteristics that differ substantially from those of microwave radiometers. Comparisons between occultation and microwave records help distinguish broad atmospheric changes from instrument-specific drift.
Uncertainty in a satellite temperature record includes radiometric calibration error, sampling error, retrieval-model error, and structural uncertainty arising from alternative processing choices. Structural uncertainty becomes visible when technically defensible data sets derived from the same raw observations produce different trends. It is not represented completely by the statistical standard error of a fitted trend line.
Spatial averaging reduces random weather variability but does not remove systematic errors shared across an instrument or satellite. Likewise, the large number of individual observations does not by itself determine the accuracy of the final climate trend, because many observations depend on the same calibration model. Long-term reliability instead derives from stable references, overlap among instruments, comparison with independent observations, and explicit propagation of processing uncertainty.
Interpretation in climate analysis
Satellite records display large short-term variations associated with the El Niño–Southern Oscillation, volcanic eruptions, and changes in atmospheric circulation. Tropospheric temperatures typically rise during major El Niño events as heat is transferred from the tropical ocean to the atmosphere. Large volcanic eruptions cool the troposphere by reducing incoming solar radiation while warming parts of the stratosphere through absorption by volcanic aerosol.
Multidecadal microwave records show warming in the troposphere and cooling in the stratosphere. This vertical pattern is physically consistent with increased concentrations of greenhouse gases together with changes in stratospheric ozone and volcanic forcing. The magnitude of a reported trend depends on the selected layer, geographical domain, time interval, and data-set version.
Comparisons with climate-model output require application of the satellite weighting functions to the simulated atmosphere. A model temperature taken from a single pressure level is not directly equivalent to a satellite product whose radiance integrates contributions across several kilometres of altitude. Consistent comparison also requires equivalent spatial coverage and temporal sampling because missing observations and orbital sampling affect monthly means.
Satellite measurements complement rather than duplicate instrumental surface temperature records. Surface analyses represent air temperature over land and a specified near-surface ocean temperature, whereas microwave satellite products represent thick atmospheric layers. Differences between their short-term behavior therefore reflect both measurement uncertainty and genuine distinctions among the physical quantities being observed.