Modeling downwelling longwave irradiance at daily resolution under all skies: A cloud-free approach

Downwelling longwave irradiance (LW_↓) is challenging and expensive to measure and is often estimated indirectly with parametric modeling of routinely measured surface-level meteorological variables.  Modeling LW_↓ under all-sky conditions typically involves “correcting” a clear- (or non-overcast) sky model estimate using solar-irradiance-based proxies of cloud cover in lieu of actual cloud cover given uncertainties and measurement challenges of the latter.  While such approaches are deemed sound, their application in time and space is inherently limited.  Here, we present a correction model free of cloud variables applicable at the true daily (24-hr.) and global scale that – irrespective of the underlying clear-sky model – yields errors over land that are lower than those from stand-alone models and on par with daytime errors from the prevailing solar-based cloud proxy corrections (rRMSD = ~7%; rMAD = ~5.5%).  We document and critically assess its performance over land and ocean independently, as well as in high elevation and cold environments representing two notoriously challenging conditions.  The cloud-free correction is found to perform better than stand-alone approaches at all subsets; however, within-subset performance differences were evident and attributable to the underlying clear-sky model, reinforcing previous findings surrounding performance thresholds of parametric models with globally-tuned parameters.