Using MODIS land surface temperatures and the Crocus snow model to understand the warm bias of ERA-Interim reanalyses at the surface in Antarctica

摘要

Moderate-Resolution Imaging spectroradiometer (MODIS) land surfacetemperatures in Antarctica were processed in order to produce a gridded dataset at 25 km resolution, spanning the period 2000–2011 at an hourlytime step. The Aqua and Terra orbits and MODIS swath width, combined withfrequent clear-sky conditions, lead to very high availability ofquality-controlled observations: on average, hourly data are available 14 h per day at the grid points around the South Pole and more than 9 h over alarge area of the Antarctic Plateau. Processed MODIS land surfacetemperatures, referred to hereinafter as MODIS T_s values, were compared with insitu hourly measurements of surface temperature collected over the entirety of the year 2009 by seven stations from the Baseline Surface Radiation Network (BSRN)and automatic weather stations (AWSs). In spite of an occasional failure inthe detection of clouds, MODIS T_s values exhibit a good performance, with abias ranging from ?1.8 to 0.1 °C and errors ranging from 2.2 to 4.8 °C rootmean square at the five stations located on the plateau. These results show thatMODIS T_s values can be used as a precise and accurate reference to test othersurface temperature data sets. Here, we evaluate the performance of surfacetemperature in the European Centre for Medium-Range Weather Forecasts (ECMWF)reanalysis known as ERA-Interim reanalysis. During conditions detected ascloud free by MODIS, ERA-Interim shows a widespread warm bias in Antarcticain every season, ranging from +3 to +6 °C on the plateau. This confirms arecent study which showed that the largest discrepancies in 2 m airtemperature between ERA-Interim and the global temperature data set HadCRUT4compiled by the Met Office Hadley Centre and the University of East Anglia'sClimatic Research Unit occur in Antarctica. A comparison with in situ surfacetemperature shows that this bias is not strictly limited to clear-skyconditions. A detailed comparison with stand-alone simulations by the Crocussnowpack model, forced by ERA-Interim, and with the ERA-Interim/landsimulations, shows that the warm bias may be due primarily to anoverestimation of the surface turbulent fluxes in very stable conditions.Numerical experiments with Crocus show that a small change in theparameterization of the effects of stability on the surface exchangecoefficients can significantly impact the snow surface temperature. TheERA-Interim warm bias appears to be likely due to an overestimation of thesurface exchange coefficients under very stable conditions.