Abstract The Advanced Microwave Sounding Unit (AMSU) images display strong dependence on the scanning angle because of the temperature gradient of the atmosphere and the change in the optical pathlength between Earth and the satellite. Using a limb-adjustment algorithm, the temperature gradients can be restored from the images. Various limb-correction algorithms have been developed for infrared and microwave sounders by aid of radiative transfer simulations. Together with the National Oceanic and Atmospheric Administration (NOAA)-16 AMSU, the NOAA-18 satellite with AMSU (launched on 20 May 2005) provides the best opportunity to collocate observations from two satellites. The collocated measurement pairs from NOAA-16 and NOAA-18 contain data for which both observations have the same scanning angle and various scanning angles—in particular, off-nadir observations from NOAA-16 and nadir observations from NOAA-18. The coincident data pair having the same scan position from NOAA-16 and NOAA-18 can be used for intercalibration of the sensors of the two satellites. The coincident data pair having nadir measurement from NOAA-18 and off-nadir measurement from NOAA-16 can be used for testing the limb-adjustment algorithm using pure satellite measurements. This study applies collocated measurements to evaluate the performance of the current NOAA microwave limb-correction algorithm for brightness temperatures at AMSU-A channels 5, 6, and 7 for the first time. With the limb correction, the warm core of Hurricane Katrina in 2005 can also be detected using a cross-scan sensor such as AMSU-A.