This study significantly contributes to alleviating the data scarcity of drought legacy effects on ecosystem photosynthesis in temperate deciduous forests with a novel machine learning approach. This is probably the first time that drought legacies on ecosystem carbon fluxes are quantified using eddy-covariance data. The study reported that the reduction in photosynthesis due to drought legacy effects was of comparable magnitude to the concurrent drought effects at the studied sites. This study thus emphasizes the importance of drought legacy effects and provides a novel analytical method to quantify legacy effects elsewhere.