The Challenge of Measuring Temperature from Space
Imagine trying to read a thermometer through a foggy, moving window while someone shines a flashlight on it. This is the exact challenge satellite sensors face when measuring Land Surface Temperature (LST).
Satellites do not measure temperature directly; instead, they capture Top-of-Atmosphere (TOA) radiance in the thermal infrared (TIR) spectrum. This raw signal is a complex mixture of three distinct components:
- The thermal energy emitted directly by the Earth's surface.
- The energy absorbed and re-emitted upward by atmospheric gases (upwelling radiance).
- The atmospheric energy emitted downward and reflected by the surface back into space (downwelling radiance).
To isolate the true surface temperature, scientists must solve the Radiative Transfer Equation (RTE) to remove these atmospheric distortions. As Youshui Zhang et al. and Ziqi Ren et al. demonstrate, traditional single-channel or mono-window algorithms invert this equation directly using a single thermal band. However, this approach requires highly precise, real-time atmospheric profiles—such as water vapor columns and air temperature profiles—to calculate atmospheric transmittance (Jiahao Yu; Yixuan Xue et al.). When these atmospheric sounding data are missing or inaccurate, single-channel retrievals break down.
The Split-Window Intuition
To bypass the need for real-time atmospheric sounding, remote sensing scientists developed the split-window algorithm (SWA).
Instead of relying on a single thermal band, SWAs utilize two adjacent thermal infrared bands (typically centered around 11 $\mu m$ and 12 $\mu m$). The core physical insight driving this approach is elegant: atmospheric water vapor absorbs radiation differently across these two nearby bands, whereas the underlying surface emissivity remains nearly identical.
[Earth Surface] ---> Emits Thermal Radiance
|
v
[Atmospheric Water Vapor]
Absorbs 11µm slightly LESS
Absorbs 12µm slightly MORE
|
v
[Satellite Dual-Bands]
Measures the difference (ΔT) to calculate
exact atmospheric attenuation
Because the atmosphere is more opaque in one band than the other, the difference in the brightness temperatures measured by the two bands reveals exactly how much interference the atmosphere caused. As Shuang Sun et al. and Yinhe Cheng et al. note, this differential absorption allows the algorithm to correct for atmospheric attenuation internally, drastically reducing dependency on external atmospheric models.
To translate this differential signal into a precise temperature correction, researchers simulate thousands of atmospheric states using global profile libraries and radiative transfer models (Hua Li 0005 et al.; Yu Liu et al.). These simulations generate robust, sensor-specific coefficients that map the dual-band brightness temperatures directly to an accurate LST across varying viewing geometries.
Untangling Temperature and Emissivity
Even with a perfect atmospheric correction, a second fundamental ambiguity remains: the temperature-emissivity separation problem. A highly emissive surface (like dense forest) at a lower temperature can emit the exact same amount of radiance as a less emissive surface (like dry asphalt) at a higher temperature.
To resolve this ambiguity, split-window algorithms must accurately estimate Land Surface Emissivity (LSE) before calculating temperature. A common strategy leverages the predictable relationship between vegetation cover and thermal emission. For instance, Kyungil Lee et al. and Manjula Ranagalage et al. integrate fractional vegetation cover derived from visible and near-infrared bands to dynamically adjust emissivity maps.
However, complex terrains and shadows often introduce severe spectral mixing errors. To combat this, Rongfang Xin et al. introduced the Normalized Shade Vegetation Index (NSVI), which isolates sunlit from shadowed vegetation to prevent emissivity biases from propagating into the final LST calculation. For global operational pipelines, researchers increasingly bypass classification-based maps entirely, favoring advanced datasets like the ASTER Global Emissivity Dataset (GED) to provide highly accurate baseline maps over complex surface types like deserts (Hua Li 0005 et al.).
Overcoming the Cloud Barrier
While split-window algorithms excel under clear skies, thermal infrared radiation cannot penetrate clouds, leaving massive data gaps during overcast conditions. To achieve true "all-sky" monitoring, the field has expanded to combine infrared insights with alternative spectrums.
Donglian Sun et al. developed multi-channel algorithms utilizing passive microwave observations. Because microwave radiation easily penetrates cloud cover, integrating these observations with high-resolution thermal data allows scientists to reconstruct continuous, daily LST fields regardless of weather conditions, fulfilling the ultimate goal of robust, uninterrupted global thermal monitoring.
Key Foundational Work
Surface Urban Heat Island Across 419 Global Big Cities (2011) — contributed 1 concept elements to this direction.
Identifying clouds over the Pierre Auger Observatory using infrared satellite data (2013) — contributed 1 concept elements to this direction.