River networks are expected to respond to global warming in a variety of ways, ranging from more extreme annual low flows to more frequent and intense flood events. Though understanding the drivers of such shifts can help inform future modeling efforts, little focus has been placed on assessing the relative roles of the atmospheric (radiative) and land-based (plant physiological) responses to climate change in altering streamflow characteristics. Here, we begin to fill this gap through a series of downscaled Earth System Model experiments that isolate the two response pathways to a quadrupling of pre-industrial CO₂ that is applied only to the atmosphere or only to the land. We show here that the physiological response to global warming (i.e., reduced stomatal conductance leading to higher soil moisture) rivals the strength of the radiative response (i.e., intensification of precipitation) in terms of increasing the frequency of the 100-year flood. On shorter timescales, we find that this plant response can also be the primary driver of increased annual mean and maximum flows for a number of basins globally, including the Amazon, Mekong, and Yangtze. These findings not only suggest the importance of including interactive plant responses to CO₂ in future flood projections, but also highlight regions, typically those with high leaf area index and low snowfall amounts, where future observational efforts should be focused to better constrain the strength of the stomatal closure effect.