A scene-mapped framework for deriving clear-sky surface longwave downward radiation from remotely sensed measurements

Accurate estimation of clear-sky surface longwave downward radiation (LWDR) is fundamental to surface energy budget studies, yet widely used parameterizations and conventional lookup-table (LUT) searches often suffer from regional biases and limited transferability. This study develops a scene-mapped LUT framework that operates in brightness-temperature (BT) space and uses surface temperature (ST) and total column water vapor (TCWV) as physical descriptors to define scene-specific subsets and constrain spectral matching to locally consistent surface–atmosphere states. Two implementations are proposed within the scene-mapped framework: S-Poly (polynomial regression) and S-LUT (sub-LUT distance search). Validation against multi-network ground measurements using MODIS inputs yielded RMSEs of 26.36 and 26.89 W m −2 and biases of −0.86 and −2.75 W m −2 for S-LUT and S-Poly, respectively. The corresponding RMSEs were lower than those of the global-LUT benchmark (29.26 W m −2 ) and the evaluated general clear-sky parameterization benchmark (30.03 W m −2 ), representing reductions of 12.2% and 10.5% relative to the latter. Both methods generally maintained stable accuracy across diverse surface types and climatic regions. S-LUT was more consistent across the evaluated configurations and achieved an RMSE of 22.48 W m −2 under the mixed MODIS–ERA5 configuration. The flexible scene-mapping framework may also provide a useful methodological concept for adapting radiation retrieval to different sensors and exploring future extensions toward all-sky conditions.