Abstract. Quantifying soil organic carbon stocks (SOC) and their dynamics accurately is crucial for better predictions of climate change feedbacks within the atmosphere-vegetationsoil system. However, the components, environmental responses and controls of the soil CO 2 efflux (R s ) are still unclear and limited by field data availability. The objectives of this study were (1) to quantify the contribution of the various R s components, specifically its mycorrhizal component, (2) to determine their temporal variability, and (3) to establish their environmental responses and dependence on gross primary productivity (GPP). In a temperate deciduous oak forest in south east England hourly soil and ecosystem CO 2 fluxes over four years were measured using automated soil chambers and eddy covariance techniques. Mesh-bag and steel collar soil chamber treatments prevented root or both root and mycorrhizal hyphal in-growth, respectively, to allow separation of heterotrophic (R h ) and autotrophic (R a ) soil CO 2 fluxes and the R a components, roots (R r ) and mycorrhizal hyphae (R m ).Annual cumulative R s values were very similar between years (740 ± 43 g C m −2 yr −1 ) with an average flux of 2.0 ± 0.3 µmol CO 2 m −2 s −1 , but R s components varied. On average, annual R r , R m and R h fluxes contributed 38, 18 and 44 %, respectively, showing a large R a contribution (56 %) with a considerable R m component varying seasonally. Soil temperature largely explained the daily variation of R s (R 2 = 0.81), mostly because of strong responses by R h (R 2 = 0.65) and less so for R r (R 2 = 0.41) and R m (R 2 = 0.18). Time series analysis revealed strong daily periodicities for R s and R r , whilst R m was dominated by seasonal (∼150 days), and R h by annual periodicities. Wavelet coherence analysis revealed that R r and R m were related to short-term (daily) GPP changes, but for R m there was a strong relationship with GPP over much longer (weekly to monthly) periods and notably during periods of low R r . The need to include individual R s components in C flux models is discussed, in particular, the need to represent the linkage between GPP and R a components, in addition to temperature responses for each component. The potential consequences of these findings for understanding the limitations for long-term forest C sequestration are highlighted, as GPP via root-derived C including R m seems to function as a C "overflow tap", with implications on the turnover of SOC.