Abstract: The Microwave Sounding Unit(MSU)on board the National Oceanic and Atmospheric Administration(NOAA)polar-orbiting satellites measures the atmospheric temperature from the surface to the lower stratosphere under all weather conditions, excluding precipitation. These instruments are extensively used to determine the atmospheric temperature trend. However, calibration errors are a major source of uncertainties in the trend determination. Recently, NOAA/National Environmental Satellite, Data and Information Services has developed a non-linear sequential calibration method using simultaneous nadir overpass(SNO)to reduce calibration errors and then a long term MSU deep layer atmospheric temperature dataset is generated based on the new calibration. The dataset is introduced, which includes global 5-day averaged deep-layer temperatures for the mid-troposphere, tropopause, and lower-stratosphere with grid resolution of 2.5° latitude by 2.5° longitude. Also, the methodologies for the dataset generation are described, which including cross-calibration, incident angle correction, grid data generation, and bias correction in gridded data. The 20-year climate trends during 1987—2006 for the three layers are obtained from the dataset. Finally, the website for acquiring the dataset is provided.