Detector-dependent radiometric correction

Each 1-km MODIS bands is compised of 10 independent detectors, and each detector must be independently calibrated. The primary mechanism to remove relative differences between detectors is by way of the Solar Diffuser (SD) calibration. However, the SD calibration seems to leave some residual differences between detectors, resulting in low-level striping which is evident in the derived product images (e.g., nLw_412).

MODIS/Aqua detector-to-detector relative calibration differences were quantified by MCST using the Moon as a reference point. For every lunar observation, integrated lunar irradiance was computed for each band and detector. Then, its value was normalized by the band averaged lunar irradiance. The residuals from this normalization were detector dependent and exhibited noisy behavior through time. Averaging of the residuals through time provided the detector-to-detector relative calibration factors.

The Ocean Biology Processing Group (OBPG) independently confirmed the MCST results. For each lunar calibration, integrated lunar reflectance was calculated for each band and detector. The nearest in time set of solar diffuser calibration factors, m1's, was used to convert the detector counts to reflectance. The focal plane temperature corrections were also applied to the data. The relative response of each detector was obtained by normalizing the reflectance for each detector by the band-averaged reflectance. The detector-dependent result was then averaged over the 29 lunar calibrations acquired to date over the Aqua mission.

Unfortunatley, the detector-to-detector analysis using lunar observation can only be performed for MODIS reflective solar bands which do not saturate when viewing the Moon. The non-saturating 1km bands are 8-12 and 17-19. Detector dependence in red through near-infrared ocean color bands, 13-16, cannot be evaluated using the lunar measurements. An alternative approach to detector-to-detector relative calibration was therefore investigated which applies to all ocean color bands and whose initial results for bands 8-12 agree relatively well with the MCST and OBPG numbers based on lunar observations. The figure below shows the comparison of detector relative calibration change obtained by MCST (red stars), OBPG (blue stars), and the alternative analysis based on sensor's top-of-the-atmosphere (TOA) radiances (circles, filled circles for the mirror side 1 and unfilled circles for the mirror side 2).

MODIS detector striping was alternatively quantified using sensor's TOA radiances (Ewa Kwiatkowska). For the analysis, runs of consecutive 20 pixels in the flight direction were extracted from processed level 2 granules. The 20-pixel runs passed through strict exclusion criteria, where

  • chlorophyll-a concentration within each run was limited to 0.15 mg/m^3
  • and a very clear atmosphere was selected with aerosol optical thickness at the 869 nm band below 0.05

  • For the extracted runs, TOA radiances at all MODIS ocean color bands were investigated and their statistical dependence on the detector and mirror side were quantified.

    Because MODIS consecutive detectors view the ocean using slightly different geometry, some detector dependence in TOA radiances is expected. The problem therefore reduced to separating instrument artifacts from atmospheric and ocean variabilities caused by viewing path differences. To apply the viewing-path correction, for each 20-pixel run additional information was extracted, including molecular (Rayleigh) radiances, aerosol radiances, and water-leaving radiances multiplied by the diffuse transmittance through the atmosphere. The contribution of these radiances to the total TOA radiance was corrected for the detector-to-detector variability using the estimated change for each geometry derived from spatially flat fields of molecular atmosphere, aerosol optical thickness, and normalized water-leaving radiances.

    The striping analysis was performed using MODIS granules from a single day, 14 August 2002. The striping correction derived from this single day is applied in the current calibration table used in the MODIS-Aqua reprocessing 1.1. More intensive analyses are planned for the future when the striping code is implemented in the ocean color processing environment. Large global datasets will then be used and possible temporal changes in the detector and mirror side behavior will be traced over the mission lifetime.

    The images below illustrate the impact on the chlorophyll-a product from the forward stream. The image on top was acquired on June 10th, 2005, and processed with the uncorrected calibration, the image below was acquired on June 13th, 2005 and processed with the corrected calibration. The same color bar is used for both images. Although some stripes are still apparent, their frequency and intensity has been reduced.

    Processed without striping correction:

    Processed with striping correction: