Robustness of Radiative Kernel Methods in Reproducing Arctic Outgoing Longwave Radiation Variability

Radiative kernel methods support process level climate analysis, yet their performance in the rapidly warming Arctic has not been systematically evaluated. We compute temperature and water vapor radiative kernels with the NASA‐GISS radiative transfer code and combine them with existing NASA‐GISS cloud histogram kernels to form a self‐consistent framework. Using the new NASA‐GISS kernel set, together with 10 additional independent kernels and three reanalyses, we reconstruct monthly outgoing longwave radiation (OLR) anomalies over 2003–2022 and evaluate them against coincident CERES satellite observations to test whether kernel methods can reproduce observed Arctic OLR variability. We find that kernel derived OLR closely matches CERES observations across all kernel and reanalysis combinations, with strong correlations between observed and kernel based OLR time series. Local temperature, water vapor, and cloud changes together explain over 90% of interannual Arctic OLR variance, indicating robust closure, though some spread remains among kernel‐reanalysis combinations. These results indicates that Arctic OLR variability is expressed through changes in the local atmospheric column, with poleward energy transport influencing OLR only indirectly through its impact on atmospheric column. Compared to the global mean, Arctic OLR variability is more strongly controlled by air temperature (>50% of variance) and less by clouds (∼13%). In contrast, OLR trend attribution to temperature and water vapor remains sensitive to kernel and reanalysis choices. Overall, this study provides a systematic validation of radiative kernel methods in the Arctic and clarifies the processes controlling present‐day Arctic OLR variability.