The evaluation of Arctic reanalyses and climate models using downwelling longwave fluxes at Summit Station, Greenland

Recent studies have shown that the downwelling longwave flux (DLW) is an important parameter in the Arctic surface energy budget. Changes in the DLW play an important role in sea ice melting over the Arctic Ocean and in surface melting over the Greenland Ice Sheet (GIS). Ground-based observations provide the most accurate estimates of the DLW, but the network of observations is sparse and records are short, so it is necessary to include estimates of DLW from gridded data sets, such as reanalyses, for a more comprehensive analysis. Thus, it is important to conduct studies evaluating both the performance of the reanalyses and the complexities involved in comparing gridded data to point observations. Two independent measurements of the DLW from ground-based observations are available at Summit Station, Greenland with an overlap period from July 2010 through August 2012; one from a Broadband Surface Radiation Network station and another derived from spectral radiances obtained by an infrared spectrometer that is part of the Integrated Characterization of Energy, Clouds, Atmospheric state and Precipitation at Summit (ICECAPS) observatory. The availability of these measurements, the homogeneity of the central GIS surface, and the relatively large distance from data assimilation locations make Summit Station an ideal location for conducting reanalysis evaluations. In this study, we use a time-frequency signal decomposition technique (e.g., wavelet analysis) to evaluate reanalysis estimates of DLW at temporal scales ranging from three hours through the annual cycle. First, we evaluate the ground-based observations of DLW at Summit Station to set context for our ability to measure this parameter. Then we use the ground-based observations to evaluate the performance of the reanalysis products and to gain insight into comparisons between point measurements and gridded data sets. Finally, we extend this analysis to DLW from CMIP5 models to evaluate the representation of modes of variability, such as the diurnal cycle, in these data sets with respect to model physics and grid size.