Conditioning-controlled retrieval of broadband land surface temperature and emissivity from paired ground-based longwave irradiance measurements

Abstract. Accurate retrieval of land surface temperature (LST) from broadband longwave radiometric measurements is fundamentally limited by the nonlinear coupling between surface emissivity and temperature, which can render the inverse problem weakly observable under low irradiance contrast. We present a conditioning-controlled retrieval methodology that estimates broadband surface emissivity and LST directly from paired ground-based upwelling and downwelling longwave irradiance measurements acquired at high temporal resolution. The approach combines adaptive temporal pairing constrained by a quasi-steady apparent surface temperature criterion with a fixed-iteration Newton inversion, and explicitly diagnoses inversion stability through Jacobian strength, residual magnitude, and observed convergence order. A formal uncertainty propagation framework is developed for both independent and correlated irradiance error structures, enabling decomposition of irradiance-driven and emissivity-driven temperature uncertainty. The method is evaluated using 39 datasets from four Surface Radiation Budget (SURFRAD) Network sites spanning diverse atmospheric conditions. The Newton inversion exhibited stable and well-conditioned behaviour across all cases, and retrieved surface temperatures agreed with independent in-situ measurements with a root mean square error of 0.54 K and a mean absolute error of 0.47 K, consistent with propagated uncertainty estimates. Results demonstrate that reliable broadband LST retrieval can be achieved without externally prescribed emissivity products when inversion conditioning and measurement uncertainty are explicitly incorporated into the retrieval design.