Molecular hydrogen (H₂) is an atmospheric trace gas with a large microbe-mediated soil sink, yet cycling of this compound throughout ecosystems is poorly understood. Measurements of the sources and sinks of H₂ in various ecosystems are sparse, resulting in large uncertainties in the global H₂ budget. Constraining the H₂ cycle is critical to understanding its role in atmospheric chemistry and climate. We measured H₂ fluxes at high frequency in a temperate mixed deciduous forest for 15 months using a tower-based flux-gradient approach to determine both the soil-atmosphere and the net ecosystem flux of H₂. We found that Harvard Forest is a net H₂ sink (−1.4 ± 1.1 kg H₂ ha⁻¹) with soils as the dominant H₂ sink (−2.0 ± 1.0 kg H₂ ha⁻¹) and aboveground canopy emissions as the dominant H₂ source (+0.6 ± 0.8 kg H₂ ha⁻¹). Aboveground emissions of H₂ were an unexpected and substantial component of the ecosystem H₂ flux, reducing net ecosystem uptake by 30% of that calculated from soil uptake alone. Soil uptake was highly seasonal (July maximum, February minimum), positively correlated with soil temperature and negatively correlated with environmental variables relevant to diffusion into soils (i.e., soil moisture, snow depth, snow density). Soil microbial H₂ uptake was correlated with rhizosphere respiration rates (r = 0.8, P < 0.001), and H₂ metabolism yielded up to 2% of the energy gleaned by microbes from carbon substrate respiration. Here, we elucidate key processes controlling the biosphere–atmosphere exchange of H₂ and raise new questions regarding the role of aboveground biomass as a source of atmospheric H₂ and mechanisms linking soil H₂ and carbon cycling. Results from this study should be incorporated into modeling efforts to predict the response of the H₂ soil sink to changes in anthropogenic H₂ emissions and shifting soil conditions with climate and land-use change.