Surface energetics and snow ablation were examined during the 1999 snowmelt season in a mountain subarctic tundra valley in the Yukon Territory of Canada. Considerations of melt energetics at small scales were made with respect to the frame of reference of the sloping surface snowpack. During relatively warm and sunny conditions early in melt, snow ablation rates were dramatically higher on the south-facing slope and strongly reduced on the north-facing slope, compared to the valley bottom. Negative spatial covariances developed between maximum snow accumulation and ablation rate during early and middle melt, with the highest ablation rates occurring on slopes with the shallowest snowpacks. Atmospheric conditions were sufficiently well mixed across the valley that reference level air temperatures and humidity among the slopes were close to levels of measurement accuracy. However, under high levels of April insolation, notable differences in incoming solar radiation to varying slopes/aspects caused relatively larger differences in net radiation and surface temperature, which were progressively magnified as shrubs and soil became exposed during snow ablation. Under cloudier conditions later in melt, the south-facing snowpack had mostly ablated, vegetation was exposed at all sites, and ablation rates were virtually identical between the valley bottom and north-facing slope. Driven primarily by initial differences in insolation and snow accumulation, surface energy fluxes changed sign and magnitude over space, not only with insolation, vegetation cover, slope, and aspect, but also with the snow cover state and ground/vegetation exposure. Melt rate was, hence, controlled by both incoming energy and evolving and initial snow states. For these reasons, and because of the slope-based frame of reference necessary to precisely define the snowmelt energy balance, simple aggregate representations of melt in subarctic mountain environments that are based on averaged energy flux, snow state, and flat-plane conceptions may require substantive corrections that should be explored in modeling studies.