The arctic ice pack is a mixture of ice of many different thicknesses. Ice growth and heat exchange are strongly influenced by thickness, particularly when the ice is thin. For this reason measurements at a particular location do not necessarily represent conditions elsewhere within the region. Large-scale heat exchange estimates must take into account contributions made by different thicknesses of ice. This requires information on the relative area covered by ice of any given thickness and on how each flux varies with thickness. Strain histories derived from the motions of several buoy and drifting station arrays were combined with climatological data on air temperatures and incoming radiation to estimate time dependent changes in the distribution of ice thickness in the Central Arctic. Thermodynamic ice models were used to determine the dependence of heat exchange and ice production on ice thickness. Large-scale fluxes were obtained by summing the area weighted contributions made by each ice thickness category. Differences between the large-scale fluxes and those based on local measurements over perennial ice were due almost entirely to the effects of young ice less than a meter in thickness. Net annual ice production in areas of thin ice totaled about 1 m when averaged over the entire area of the strain array. In contrast to the very small annual values measured over multiyear ice, large-scale turbulent hat losses were close to 200 MJ m-2 year-1, similar in magnitude to the net radiation. Absorption of shortwave radiation by summer leads resulted in annual net radiation totals for the region which were more than double those over the ice. Solar energy absorbed in the water played a major role in the mass balance of the ice cover. Intermediate thicknesses (0.2--0.8 m) of young ice, rather than open leads, exerted the greatest influence on ice production, heat input to the atmosphere, and salt input to the ocean. Monthly and annual heat flux totals obtained with different strain histories showed little correlation with the average divergence, suggesting that the variability of the strain field may be more important than the long-term average of the strain components.