Forests in the south-eastern United States experienced a prolonged dry spell and above-normal temperatures during the 1995 growing season. During this episode, nearly continuous, eddy covariance measurements of carbon dioxide and water vapour fluxes were acquired over a temperate, hardwood forest. These data are used to examine how environmental factors and accumulating soil moisture deficits affected the diurnal pattern and magnitude of canopy-scale carbon dioxide and water vapour fluxes. The field data are also used to test an integrative leaf-to-canopy scaling model (CANOAK), which uses micrometeorological and physiological theory, to calculate mass and energy fluxes. When soil moisture was ample in the spring, peak rates of net ecosystem CO2 exchange (NF) occurred around midday and exceeded 20 μmol m−2 s−1. Rates of NK were near optimal when air temperature ranged between 22 and 25°C. The accumulation of soil moisture deficits and a co-occurrence of high temperatures caused peak rates of daytime carbon dioxide uptake to occur earlier in the morning. High air temperatures and soil moisture deficits were also correlated with a dramatic reduction in the magnitude of NE. On average, the magnitude of NE decreased from 20 to 7 μmol m−2 s−1 as air temperature increased from 24 to 30°C and the soil dried. The CANAOK model yielded accurate estimates of canopy-scale carbon dioxide and water vapour fluxes when the forest had an ample supply of soil moisture. During the drought and heat spell, a cumulative drought index was needed to adjust the proportionality constant of the stomatal conductance model to yield accurate estimates of canopy CO2 exchange. The adoption of the drought index also enabled the CANOAK model to give improved estimates of evaporation until midday. On the other hand, the scheme failed to yield accurate estimates of evaporation during the afternoon.