An inflight calibration for AVHRR visible and near-IR bands is discussed and applied to NOAA-7, NOAA-9 and NOAA-11 from 1981 to 1990. The approach, independent of ground support, relies on three unique earth-atmosphere phenomena: molecular scattering over the ocean for absolute visible band calibration, ocean glint to transfer the calibration from the visible band to the near-IR band, and desert reflectance to monitor, independently, the stability of the visible and near-IR bands. The resulting two calibration methods differ in the brightness range and spectral response of the radiance source (molecular scattering versus desert reflectance). Both methods agreed on calibration deterioration for NOAA-9 of 10 ± 3 per cent in the visible band and 16 ± 2 per cent in the near-IR shortly after launch, and 22 ± 2 per cent in both bands three years later. The ocean method showed an increase of 6 per cent and 9 per cent in the sensitivity of the visible and near-IR bands respectively over the lifetime of NOAA-7. However, the desert method showed a decrease of 10 and 12 per cent, respectively, in the two bands. Possible reasons for the differences between the resultant calibration for NOAA-7 from the two methods are discussed. NOAA-11 was calibrated using the desert radiances only. Its inflight response was lower from the preflight response by 21 per cent in channel 1 and 33 per cent in channel 2, a short time after launch. Recommendations for the absolute calibration of NOAA-7, -9 and -11 are given for channels 1 and 2 as a function of time and are compared with other published AVHRR calibrations. The calibration deterioration results in a change in the vegetation index (NDVI) between 0.0 and 0.09. A simple correction scheme is suggested for uncorrected NOAA-7, -9 and -II NDVI and implications of the calibration ratio on AVHRR remote sensing are discussed.