For several years, the Arctic has now been in the focus of scientific debate on climate change. It is a region of high spatio-temporal climate variability and additionally a region of high climate sensitivity due to strong feedback processes like the ice-albedo feedback. As there is a strong linkage to the global climate system, changes in the Arctic impact onto the global climate. In modeling the Arctic climate, regional climate models are an important tool because with their high resolution they provide the possibility to account for the horizontally heterogeneous soil and surface characteristics. Here, the regional climate model HIRHAM is used with 25km resolution for the analysis of spatial patterns, variability and trends of the Arctic climate and temperature-derived indices describing climate extremes. Inter-annual temperature variability (ITV) for present-day conditions (1958 to 2008) is examined from station data, the ERA40 re-analysis and HIRHAM results. It shows a pronounced decadal variability and specific regional and seasonal characteristics. Seasonal temperatures in general show warming trends, though they are mostly not statistically significant. Intra-seasonal extreme temperature range (ETR) trends were of mixed sign and only significant from station data over the eastern Russian Arctic. In general, the spatial pattern and magnitude of the Arctic temperature variability, both of seasonal temperature and intra-seasonal ETR, are well reproduced by HIRHAM. The large variability of the Arctic temperature, which is inherent in the analysis period, demonstrates that natural variability is an important factor in the Arctic climate. This variability is not restricted to climate means but also appears in temperature extremes. An analysis of station-derived an re-analysis-based climate indices shows complex behavior, some measures like frost days show consistent decrease (i.e. warming), while others like cold spell days provide a more diverse picture. As with seasonal temperatures, only few trends are found statistically significant. The indices examined exhibit strong inter-annual and decadal-scale variability and heterogeneous spatial patterns. These climate indices are then employed in the validation of the HIRHAM model. The model well reproduces trends and variability of most indices while there is an offset in some absolute values (e.g. frost days, growing degree days). Other measures like cold and warm spells are calculated with non-systematic biases; deviations in trends and variability occur in summer for cold spells and in spring and summer for warm spells due to an earlier spring warming and a too low variability of the maximum temperature over sea ice in HIRHAM. HIRHAM is furthermore used as a downscaling tool for future projections (ECHAM5/MPIOM output under the IPCC scenario SRES A1B). The strong increase in mean annual air temperature (5–8 K) is expected to increase active layer thickness and permafrost boundaries will move northwards. On top of this general warming trend, the additional analysis of future changes, using the mean conditions for the warmer climate, highlights some particularly vulnerable regions (West Siberian Plain, Laptev Sea coast, Canadian Archipelago), which are projected to be warmer, to experience increased warm spells and to be wetter in summer; all this contributes to amplify the permafrost degradation initiated by the general warming. Different realizations of HIRHAM are run for a sensitivity study looking into the importance of land-surface-conditions for climate model projections. The different model setups are: (1) the incorporation of freezing/thawing of soil moisture, (2) the inclusion of top organic soil horizons typical for the Arctic and (3) a vegetation shift due to a changing climate. Direct thermal responses in 2m air temperature and turbulent heat fluxes over land lead to changes in mean sea level pressure and geopotential height throughout the Arctic. This points to the importance of dynamical feedbacks within the atmosphere-land system. Land and soil processes have a distinct remote influence on large scale circulation patterns in addition to their direct, regional effects. The projected changes are clearly afflicted with uncertainties due to the different setups for land-surface-conditions; the highest temperature uncertainties are found over tundra regions. This demonstrates that for an improvement of the land-surface scheme of the HIRHAM model, all three representations of land-surface-processes have to be incorporated.