Abstract The role of high-frequency and low-frequency eddies in the melt onset of Arctic sea ice is investigated through an examination of eddy effects on lower-tropospheric (1000–500 hPa) meridional heat transport into the Arctic and local surface downwelling shortwave and longwave radiation. Total and eddy components of the meridional heat transport into the Arctic from 1979 to 2012 are calculated from reanalysis data, and surface radiation data are acquired from the NASA Clouds and the Earth’s Radiant Energy System (CERES) project dataset. There is a significant positive correlation between the mean initial melt date and the September sea ice minimum extent, with each quantity characterized by a negative trend. Spatially, the earlier mean melt onset date is primarily found in a region bounded by 90°E and 130°W. The decline in this region is steplike and not associated with an increase in meridional heat transport but with an earlier appearance of above-freezing temperatures in the troposphere. In most years, discrete short-duration episodes of melt onset over a large area occur. In an investigation of two of these melt episodes, a positive total meridional heat transport is associated with the peak melt, with the product of high-frequency eddy wind and mean temperature fields being the most important contributor. Additionally, there is a key positive anomaly in surface downwelling longwave radiation immediately preceding the peak melt that is associated with increased cloud cover and precipitable water. These results suggest the importance of carefully considering and properly representing atmospheric eddies when modeling the melt onset of Arctic sea ice.