In this work, we analyze the two-dimensional distribution of mean and intermodel spread of Arctic sea ice and climate change at the time of CO2 doubling and their connection using the simulations from the second phase of the Coupled Model Intercomparison Project (CMIP2). Arctic surface warming at the time of CO2 doubling is found to be not evenly distributed and ranges from 1¿ to 5¿C. The intermodel spread is pronounced in the Arctic Ocean, particularly in the Barents Sea. Reduction of sea-ice thickness (SIT) is in the range 0.3--1.8 m and mainly appears in the Greenland-Barents Seas. Meanwhile, sea-ice concentrations (SIC) decrease more than 10% in most regions of the Arctic Ocean. The sensitivity of Arctic surface air temperature change with respect to sea-ice area change is model-dependent. For some models, the sensitivity is different even in different periods of the transient integration. Values of the sensitivity vary from -2.0 to -0.5¿C/106 km2 for most CMIP2 models. A colder (warmer) Arctic climate may favor a higher (lower) sensitivity. The simulated mean and intermodel spread patterns of surface air temperature (SAT) change are similar to those of SIT and sea level pressure (SLP) changes. This implies that the mean and intermodel spread of projected Arctic climate change are influenced by the interaction between sea ice and the atmosphere. Both SIT and SIC are sensitive to the increase in greenhouse gas concentrations, and are connected with SAT and SLP changes in the Arctic. The average of all model simulations indicates that the north-south SLP gradient and the mean westerly winds are enhanced by CO2 doubling. Finally, both the mean and intermodel spread patterns show considerable differences between models with and without flux adjustment in some regions.