Two cottonwood plantations were grown at different CO₂ concentrations at the Biosphere 2 Laboratory in Arizona to investigate the response of isoprene emission to elevated [CO₂] and its interaction with water deficits. We focused on responses due to seasonal variation and variation in the mean climate from one year to the next. In fall and in spring, isoprene emission rate showed a similar inhibition by elevated [CO₂], despite an 8–10°C seasonal difference in mean air temperature. The overall response of isoprene emission to drought was also similar for observations conducted during the spring or fall, and during the fall of two different years with an approximate 5°C difference in mean air temperature. In general, leaf-level isoprene emission rates, measured at constant temperature and photon-flux density, decreased slightly, or remained constant during drought, whereas ecosystem-level isoprene emission rates increased. The uncoupling of ecosystem- and leaf-scale responses is not due to differential dependence on leaf area index (LAI) as LAI increased only slightly, or decreased, during the drought treatments at the same time that ecosystem isoprene emission rate increased greatly. Nor does the difference in isoprene emission rate between leaves and ecosystems appear to be due solely to increases in canopy surface temperature during the drought, though some increase in temperature was observed. It is possible that still further factors, such as increased penetration of PPFD into the canopy as a result of changes in leaf angle, reduced sink strength of the soil for atmospheric isoprene, and decreases in the mean C_i of leaves, combined with the small increases in canopy surface temperature, increased the ecosystem isoprene emission rate. Whatever the causes of the differences in the leaf and ecosystem responses, we conclude that the overall shape of the leaf and ecosystem responses to drought was constant irrespective of season or year.