Abstract: It is practical significance to develop a long-term, high-accuracy dataset of the surface heat sources (including surface sensible heat flux (SH) and latent heat flux (LE)) in Qingzang plateau, which can meet the time requirement of climate monitoring and prediction. Among various atmospheric reanalysis data, the NCEP-II atmospheric reanalysis data can meet the operational time requirement of climate monitoring and prediction. However, the NCEP-II SH and LE calculated based on the classic Monin-Obukhov similarity theory have large errors. Previous studies have developed SH and LE datasets in the Qingzang plateau using the Maximum Entropy Production (MEP) model and the multi-source datasets including satellite remote sensing data and atmospheric reanalysis data. However, these SH and LE datasets are short in time and cannot meet the timeliness requirements. In this paper, using the MEP model and daily average net radiation, surface temperature, and soil moisture data extracted from the NCEP-II atmospheric reanalysis dataset, we establish SH and LE in the Qingzang plateau area from 1980 to 2023. We verify the reliability of the new dataset by using the Qingzang plateau intensive observation data and analyze the temporal and spatial variation characteristics of SH and LE. The result shows that the correlation coefficients between monthly mean MEP SH (SH_MEP) and LE (LE_MEP) and the observations are 0.93 and 0.82, respectively, and the root mean square errors（RMSE） are 11.91 and 13.80 W/m², respectively. The results are better than those of the ERA5, ERA-Interim, MERRA-2, JRA-55, and NCEP-II atmospheric reanalysis data, and can meet the timeliness and quality needs of climate monitoring and prediction operations. Moreover, for the period 1980—2023, summer SH_MEP exhibits a significant downward trend, with a linear trend of -0.48 W/(m²·10 a), while LE_MEP shows an unsignificant upward trend.