Estimate of noise and systematic error in early thermal infrared data of the Moderate Resolution Imaging Spectroradiometer (MODIS)

The present article is a study of the channel-dependent noise and systematic error in Moderate Resolution Imaging Spectroradiometer (MODIS) thermal infrared (TIR) data in bands 20–25 and 27–36 using early MODIS data over lake and ocean sites in clear-sky days acquired with the A-side of scan mirror and electronics before the end of October 2000. In 14 cases of subarea sites with a size of 10 lines by 16 pixels each line, where the brightness temperature in band 31 changes within ±0.1 K, average and standard deviation values of brightness temperatures in 10 channels (consisting a 10-element linear detector array) of 16 MODIS TIR bands show the channel-dependent noise and systematic errors. A method is presented to separate the channel-dependent systematic error from noise due to variations in surface and atmospheric conditions. This case study shows that in the total 160 MODIS TIR channels, three channels are too noisy to use, 10 channels have significant channel-dependent noise, and all other channels met the specification of noise equivalent temperature difference (NEDT). After correcting the channel-dependent systematic errors, the quality of the MODIS TIR data is significantly improved in bands 22–25 and 27–30. More uniform sites in the brightness temperatures of bands 33–36 are needed to accurately estimate the values of channel-dependent systematic errors because the effects of spatial variations in atmospheric conditions are often larger in these four bands.