The regional climate model HIRHAM4 coupled with two different land surface models (ECHAM4-LSM and NCAR-LSM) has been applied over the entire Arctic region. Model results have been compared with the observed monthly soil temperature and snow water equivalent from several weather stations in east Siberia and west Russia. Our study shows that the seasonal cycle of soil temperature is reasonably well reproduced by both versions of the model. HIRHAM4-ECHAM4-LSM produces cold soil temperature bias during winter, and the use of HIRHAM4-NCAR-LSM reduces such bias because of its more advanced land surface schemes. Insulation by snow on the ground surface is one of the main controlling factors of soil temperature, and the increased association between snow and soil temperature in HIRHAM4-NCAR-LSM has led to the improvement in winter soil temperature. We further established that the influences of land surface schemes and their coupling on the atmospheric conditions not only affect climate simulations on land but also in remote Arctic Ocean regions. Similarly as for the present day climate, two different land surface schemes can have significant influences on the simulated future climate. These influences imply an additional uncertainty in future climate projections. Copyright 2006 by the American Geophysical Union.