Agricultural greenhouse gas (GHG) has a significant effect on climate change. The ability to quantify agricultural GHG and implement practices to mitigate its emissions is of great significance. Optimal GHG mitigation practices in the middle and lower reaches of the Yangtze River, China, have not been well documented. In this study, historical (2000–2014) and projected (2015–2049) methane (CH4) and nitrous oxide (N2O) emissions from paddy‐upland crop rotations were simulated at a county scale using the denitrification–decomposition model. The results demonstrated (1) historical mean emissions of CH4, N2O, and global warming potential (GWP) were 158.6 kg‐C ha−1 year−1, 0.75 kg‐N ha−1 year−1, and 22.15 Tg CO2‐eq year−1 in the study area (1 Tg = 1012 g); (2) CH4 emission rates significantly increased, while there was no significant change in N2O under projected RCP4.5 and RCP8.5 climate scenarios, and RCP8.5 induced 10% more CH4 than RCP4.5; (3) marginal flooding coupled with reduced‐tillage practice was the optimal strategy to mitigate GHG emissions due to its lowest GWP. This study would provide insight into the agricultural GHG emissions, as well as the impact on climate change. Also, a theoretical basis for policymaking would be provided by this study. © 2020 Society of Chemical Industry and John Wiley & Sons, Ltd.