Abstract: The global microwave land surface emissivity (MLSE) atlas is currently an important initial guess or simulation value in satellite data assimilation observation operators. However, due to the lack of "real" observation of MLSE, the quality and reliability of existing global MLSE products based on multiple microwave sensors are unknown, and there is also a lack of global MLSE datasets with good quality and wide frequency coverage. This study selects three sets of MLSE atlases (SSMI/S, AMSU-A/B, ATMS) used in RTTOV (Radiative Transfer for TOVS) as well as the background dataset included in the TELSEM2 MLSE estimation tool, two sets of AMSR-E datasets (AMSR-E1 and AMSR-E2), and one set of FY-3D dataset. Based on statistical analysis techniques, global spatiotemporal consistency of the seven MLSE products is evaluated. Furthermore, six typical land cover types (Amazon tropical rainforest, Northern coniferous forest, North China plain farmland, Qingzang plateau grassland, Sahara desert, and Greenland ice cover) are selected to conduct a dependency assessment for each product based on land cover type and frequency. Based on the selection of emissivity products, the TELSEM2 emissivity products are optimized, and a new set of global MLSE data with a frequency coverage of 6.925—150.0 GHz is created (named CoTELSEM2). The results show that there is a significant inversion error in the AMSR-E2 product, making it almost unusable. The MLSEs on the monthly scale of AMSR-E1, TELSEM2, SSMI/S, AMSU-A/B, ATMS and FY-3D have a good spatiotemporal consistency with an average spatial correlation coefficient (R) of 0.887—0.928, with TELSEM2 having the highest value (0.928) and FY-3D having a slightly lower value (0.914). The mean absolute deviation (MAD) is 0.031—0.037, with TELSEM2 having the lowest (0.031) and FY-3D having the highest (0.041). The spatial consistency is better with the same scanning method than with different scanning methods. The cone and cross-track scanning methods perform better with TELSEM2 and AMSU-A/B products, respectively. ATMS 51.7 GHz channel MLSE has a systematic overestimation, and 23.8 GHz and 89.0 GHz channel MLSE for AMSR-E1 and FY-3D products also have a systematic overestimation in areas with high vegetation coverage, while FY-3D MLSE shows significant overestimation or underestimation at different times, regions, and frequencies. The seasonal variability uncertainty of MLSE increases with increasing frequency. Little uncertainty is found in the emissivity of the Sahara and Amazon tropical rainforests, while the uncertainty in the northern coniferous forest is the greatest. Overall, TELSEM2 and AMSU-A/B have higher quality reliability, followed by SMMI/S and ATMS, and AMSR-E1 and FY-3D have relatively poor quality reliability. The optimized new CoTELSEM2 emissivity product shows a good dependence on land cover type and frequency, while the uncertainty of global MLSE seasonal change has a strong dependence on land cover type, of which the uncertainty of Sahara desert and Amazon tropical rainforest emissivity is very small, while the uncertainty of northern coniferous forest is the largest, and the high frequency is significantly greater than the low frequency.