We examined a 6-year record of automated chamber-based soil CO2 efflux (Fs) and the underlying processes in relation to climate and canopy gas exchange at an AmeriFlux site in a seasonally drought-stressed pine forest. Interannual variability of Fs was large (CV=17%) with a range of 427 g C m−2 yr−1around a mean annual Fs of 811 g C m−2 yr−1. On average, 76% of the variation of daily mean Fs could be quantified using an empirical model with year-specific basal respiration rate that was a linear function of tree basal area increment (BAI) and modulated by a common response to soil temperature and moisture. Interannual variability in Fs could be attributed almost equally to interannual variability in BAI (a proxy for above-ground productivity) and interannual variability in soil climate. Seasonal total Fs was twice as sensitive to soil moisture variability during the summer months compared with temperature variability during the same period and almost insensitive to the natural range of interannual variability in spring temperatures. A strong seasonality in both root respiration (Rr) and heterotrophic respiration (Rh) was observed with the fraction attributed to Rr steadily increasing from 18% in mid-March to 50% in early June through early July before dropping rapidly to 10% of Fs by mid-August. The seasonal pattern in Rr (10-day averages) was strongly linearly correlated with tree transpiration (r2=0.90, P<0.01) as measured using sap flux techniques and gross ecosystem productivity (GEP, r2=0.83, P<0.01) measured by the eddy-covariance approach. Rr increased by 0.43 g C m−2 day−1 for every 1 g C m−2 day−1 increase in GEP. The strong linear correlation of Rr to seasonal changes in GEP and transpiration combined with longer-term interannual variability in the base rate of Fs, as a linear function of BAI (r2=0.64, P=0.06), provides compelling justification for including canopy processes in future models of Fs.