Estimation of Detector Biases in MODIS Thermal Emissive Bands

The combination of multiple detectors and a double-sided mirror in the Moderate Resolution Imaging Spectroradiometer (MODIS) introduces striping in the imagery despite the use of onboard calibration systems. This results in a sharp unidirectional pattern across scan lines and leads to significant radiometric uncertainties in level-1 radiances, which further propagate into derived level-2 geophysical products such as the sea surface temperature. Quantifying and minimizing such uncertainties are the key steps in the improvement of remotely sensed data from MODIS. This paper describes a deterministic technique for the estimation of biases due to detector-to-detector miscalibration, mirror banding, and random striping. Using a unidirectional variational model previously introduced for destriping purposes, the scan line noise is first extracted from the observed scenes and then used to evaluate the thermal bias of detectors. The algorithm is applied to Terra and Aqua MODIS thermal emissive bands. Its qualitative performance is illustrated on several representative scenes, and quantitative results of detector biases are reported and discussed.