Livestock manure management accounts for almost 10% of greenhouse gas emissions from agriculture globally, and contributes an equal proportion to the US methane emission inventory. Current emissions inventories use emissions factors determined from small-scale laboratory experiments that have not been compared to field-scale measurements. We compiled published data on field-scale measurements of greenhouse gas emissions from working and research dairies and compared these to rates predicted by the IPCC Tier 2 modeling approach. Anaerobic lagoons were the largest source of methane (368 ± 193 kg CH4 hd−1 yr−1), more than three times that from enteric fermentation (~120 kg CH4 hd−1 yr−1). Corrals and solid manure piles were large sources of nitrous oxide (1.5 ± 0.8 and 1.1 ± 0.7 kg N2O hd−1 yr−1, respectively). Nitrous oxide emissions from anaerobic lagoons (0.9 ± 0.5 kg N2O hd−1 yr−1) and barns (10 ± 6 kg N2O hd−1 yr−1) were unexpectedly large. Modeled methane emissions underestimated field measurement means for most manure management practices. Modeled nitrous oxide emissions underestimated field measurement means for anaerobic lagoons and manure piles, but overestimated emissions from slurry storage. Revised emissions factors nearly doubled slurry CH4 emissions for Europe and increased N2O emissions from solid piles and lagoons in the United States by an order of magnitude. Our results suggest that current greenhouse gas emission factors generally underestimate emissions from dairy manure and highlight liquid manure systems as promising target areas for greenhouse gas mitigation.