Abstract. Winter CO2 fluxes represent an important component of the annual carbon budget in northern ecosystems. Understanding winter respiration processes and their responses to climate change is also central to our ability to assess terrestrial carbon cycle and climate feedbacks in the future. The factors influencing the spatial and temporal pattern of winter respiration (RECO) of northern ecosystems are poorly understood. For this reason, we analyzed eddy covariance flux data sets from 57 ecosystem sites ranging from ~35° N to ~70° N. Deciduous forests carry the highest winter RECO ratios (9.7–10.5 g C m−2 d−1), when winter is defined as the period during which air temperature remained below 0 °C. By contrast, wetland ecosystems had the lowest winter RECO (2.1–2.3 g C m−2 d−1). Evergreen needle-leaved forests, grasslands and croplands were characterized by intermediate winter RECO values of 7.4–7.9 g C m−2 d−1, 5.8–6.0 g C m−2 d−1, and 5.2–5.3 g C m−2 d−1, respectively. Cross site analysis showed that winter air or soil temperature, and the seasonal amplitude of the leaf area index inferred from satellite observation, which is a proxy for the amount of litter available for RECO in the subsequent winter, are the two main factors determining spatial pattern of daily mean winter RECO. Together, these two factors can explain 71% (Tair, ΔLAI) or 69% (Tsoil, ΔLAI) of the spatial variance of winter RECO across the 57 sites. The spatial temperature sensitivity of daily winter RECO was determined empirically by fitting an Arrhenius relationship to the data. The activation energy parameter of this relationship was found to decrease at increasing soil temperature at a rate of 83.1 KJ ° C-1 (r = −0.32, p < 0.05), which implies a possible dampening of the increase in winter RECO due to global warming. The interannual variability of winter RECO is better explained by soil temperature than by air temperature, likely due to the insulating effects of snow cover. The increase in winter RECO with a 1 °C warming based calculated from the spatial analysis was almost that double that calculated from the temporal analysis. Thus, models that calculate the effects of warming on RECO based only on spatial analyses could be over-estimating the impact.