The aims of the study were to identify factors related to temporal and spatial variation in forest soil CO2 efflux (Fs), compare measurement chambers, and to test effects of a climate change experiment. The study was based on four-year measurements in upland Scots pine forests. Momentary plot averages of Fs ranged from 0.04 to 1.12 gCO2m −2 h −1 and annual estimates for the forested area from1750 to 2050 gCO2m −2. Soil temperature was a dominant predictor of the temporal variation in Fs (R 2 =76-82%). A temperature and degree days model predicted Fs of independent data within 15% on the average but underestimated it during the peak efflux period (July-August), possibly because of seasonal pattern in growth of roots and mycorrhiza. A comparison sub-study indicated that the reliability of the measurement chambers was not related to the principle i.e. non-steady-state through-flow, non-steady-state non-through-flow or steady-state through-flow. Spatial variability of Fs within 400 m 2 plots in four stands was large; coefficients of variation (CV) ranged from 0.10 to 0.80, with growing season averages of 0.22-0.36. A positive spatial autocorrelation was found at short distances (3-8 m). In data from several stands, thickness of the humus layer explained 28% of the variation in Fs, and with the distance to the closest trees it explained 40%. Fs also correlated with root mass of the humus layer. Between-plot differences in Fs were small. In the climate change experiment, CO2 enrichment and air warming consistently, but not always significantly, increased Fs in whole-tree chambers. Their combined effect was additive, with no interaction; i.e. +23-37% (elevated CO2), +27-43% (elevated temperature), and +35-59% (combined treatment), depending on year. Air warming was a significant factor in the 4-year data according to ANOVA. Temperature sensitivity of Fs under the warming, however, decreased in the second year.