The objective of our study was to estimate CO2 efflux from the snow surface in a deciduous broadleaf forest throughout the duration of winter on the basis of a continuous measurement of CO2 concentration at the soil-snow interfaces. Continuous CO2 efflux was estimated from the CO2 concentration gradient across the snow cover. We examined the accuracy of the values obtained from the CO2 concentration gradient across the snow cover by comparing them with the values observed by the closed static chambers. Although each couple of the values from both methods had relatively large differences, the amounts of the values were consistent. The average CO2 efflux ¿ 1SD was 1.4 ¿ 1.0 gCO2 m-2 d-1 throughout the duration of winter. The estimated CO2 efflux in winter is considered to be reasonable, compared with the value obtained by a relationship between soil temperature and soil respiration in the same forest and with the values reported in forest ecosystems in middle latitudes. Our value is, however, smaller than the value obtained by the eddy covariance technique at a tower in the same forest. We attempted to estimate the hypothetical CO2 efflux, minimizing the impact of pressure pumping. The average CO2 efflux was estimated at 2.0 gCO2 m-2 d-1.