Modifications to the MODIS Aqua Radiometric Calibration

Summary

The main change to the MODIS Aqua radiometric calibration is that starting with the 2009 reprocessing, vicarious calibration coefficients (derived using SeaWiFS level 3 water-leaving radiances as truth fields) are being applied. Temporally varying coefficients are applied to bands 8 and 9 to eliminate a downward trend of the global average Rrs of those two bands (412nm and 443nm). For bands 10-14 (488nm to 678nm), the coefficients are applied only averaged over time. This reduces a scan angle dependency of the Rrs of those bands (which was particularly significant in bands 13 and 14), but leaves the trends derived from those bands intact. This means that for these bands, the global trends trends are still derived from the MODIS Aqua solar diffuser and lunar calibration results, they are independent of SeaWiFS. On the other hand, for 412nm and 443nm, the trends are not independent of SeaWiFS anymore, they are in fact expected to be identical. The specific algorithm and the magnitude of the modifications are described below.

The second major change to the MODIS Aqua radiometric calibration is a new algorithm from MCST to trend the response-versus-scan (RVS) of the red and NIR bands. Previously, the solar diffuser was the only calibration source that was used to calibrate these bands, because these bands saturated when viewing the moon. The new algorithm uses some of the unsaturated pixels and ratios them to an unsaturated band. The new RVS trending improved significantly the trending of the global averages of the Rrs of band 13 (667nm) and band 14 (678nm) relative to each other, which significantly reduced an erroneous trend in the globally averaged fluorescence line height product, see below.

Vicarious Calibration Method

The standard polarization correction equation for an uncalibrated instrument is 

Lm = M1 Lt + M12 (Qt cos 2alpha + Ut sin 2alpha)+ M13 (- Qt sin 2alpha + Ut cos 2alpha) + M14 V_t
where (Lt, Qt, Ut, Vt) is the Stokes vector at the top-of-atmosphere (TOA), Lm is the measured radiance, and alpha is a rotation angle to adjust for different reference frames. Since Vt is very close to zero at the TOA, M14 is an irrelevant parameter for MODIS. The parameters M12 and M13 were determined prelaunch at scan angles from -45deg to +45deg, for each band, mirror side, and detector. The variations of these parameters with detector were considered suspect and not applied in the ocean color processing (Meister et al.).

The cross calibration method has been described by Kwiatkowska et al., so here we provide only a brief summary. For a given day, the level 3 water-leaving radiances from SeaWiFS are used to predict the TOA radiances as seen by MODIS on that day, using the atmospheric correction approach from Gordon and Wang in reverse mode (Franz et al.). All components of the Stokes vector (L, Q, U, V) are modeled. This allows not only a correction for the radiometric calibration parameters M$_{11}$, but also for the polarization correction parameters m12 and m13. The modeled TOA radiances are compared to the radiances measured by MODIS for every scan angle, mirror side, and detector. This means that the instrument characterization parameters (M11, M12, M13) can be derived as a function of scan angle, mirror side, and detector. To reduce noise, the scan angle dependence is modeled by a cubic function for M11, as a linear function for M12 and M13.

The retrieval of the instrument characterization parameters is repeated for one day in every month of the mission. This results in a time series for the instrument characterization parameters. The results are smoothed over time using 5th order polynomials before they are applied in the processing of ocean color products.

MODIS Aqua radiometric correction coefficients

Fig.1 shows the results for the M11 coefficients for bands 8, 9, 10, 12, 13 (center wavelengths of 412nm, 443nm, 488nm, 547nm, and 667nm, resp.) as a function of frame (or view angle). It can be seen that the corrections required for band 8 are much larger than for the other bands, especially at the end of the mission.

E.g., at the end of the scan, Fig.1 suggests that the TOA radiances of band 8 need to be reduced by more than 3%, whereas they should be increased by 4% around frame 500. The corrections for band 9 are usually less than 1%. For bands 10-14, the corrections are so small that it was decided to average the M11 coefficients over time, but to keep the scan angle dependence. Therefore, only one line is shown for these bands in Fig.1.

The corrections are needed for the operational processing of MODIS Aqua. This is a challenge, because the crosscalibration only provides correction coefficients for past time periods. At the moment, we are applying a temporal linear extrapolation of the results, which has worked reasonably well for the first three months of 2010. Continuous monitoring is necessary to evaluate how long this approach yields sufficient results. At the moment, we are planning to derive crosscalibration coefficients on a monthly basis.

Fig.1: Correction to the MODIS calibration coefficients for bands 8-10 and 12-13 as a function of frame. The frame is proportional to the view angle, frame = 0 corresponds to the beginning of the scan (view angle of -55deg), frame = 1353 corresponds to the end of the scan (view angle of +55deg). Black/blue is for the beginning of the mission, orange/red for the end of 2009.}

The polarization correction coefficients M12 for MODIS Terra changed remarkably over the mission, by more than 0.3 for band 8 (Kwiatkowska et al.). For MODIS Aqua, there is no detectable long term trend, see Fig.2. There is a significant seasonal oscillation which is much stronger than what was found for MODIS Terra. The reason for this oscillation needs to be investigated, but since there is no indication that there is a long term trend, no correction is applied to the prelaunch polarization coefficients. The prelaunch polarization coefficients are used averaged over detectors, but they are mirror side and scan angle dependent.

Fig.2: MODIS polarization coefficients m12 = M12/M11 as a function of time. Black, blue, and red are for detectors 1, 5, 10, respectively.

Impact on MODIS Aqua ocean color products

The impact on the MODIS Aqua remote sensing reflectance at 412nm (Rrs(412)) is shown in Fig.3. The black dots show the global average over oligotrophic water for a 4-day period from each month (after removal of the global seasonal oscillation), the blue line shows a running average. The large drop in 2008 of about 15% has been removed.


Fig.3: Rrs(412) before (top) and after (bottom) correction.

The MODIS Aqua fluorescence line height product (FLH) is extremely sensitive to the difference of bands 13 and 14 (667nm and 678nm, resp.). The old operational products showed a decrease in the band 14 remote sensing reflectance over the mission, but not in band 13. This lead to a decrease with time of the FLH product of about 30%. A new lunar trending approach for bands 13-16 by the MODIS Calibration and Support Team (which uses non-saturated pixels of those bands) yielded very consistent temporal trends for the remote sensing reflectances of bands 13 and 14, which removed the trend in the FLH product, see Fig.4.


Fig.4: FLH before (top) and after (bottom) correction.

The changes to the calibration of bands 13-16 due to the new MCST algorithm are shown in Fig.5.

Fig.5: Change of the calibration of bands 13-16 due to new MCST algorithm. Same color coding as in Fig. 1.

Other improvements of the 2010 MODIS Aqua reprocessing due to calibration adjustments include

  • The new lunar calibration approach reduces a long term increase in the global angstroem coefficients.
  • Improvement of the agreement to SeaWiFS of chlorophyll-a concentrations in oligotrophic waters (a bias of 15-20% was reduced to 1-2%, at least partly due to a readjustment of the out-of-band correction for MODIS Aqua band 12, and the trends in SeaWiFS and Aqua are very similar now; the new MODIS Aqua trend is shown in Fig.6).
  • The scan angle dependence of the remote sensing reflectance of all bands has been improved (largely due to the application of the M11 coefficients shown in Fig.1).

The largest improvement was achieved for bands 8, 13, and 14, see slides 15 and 16 in this ppt. However, bands 13 and 14 still show a residual scan angle dependence in the later part of the mission.

Fig.6: Chlorophyll-a after correction.

References

Kwiatkowska, E. J., Franz, B. A., Meister, G., McClain, C. R., & Xiong, X. (2008). Cross calibration of ocean-color bands from Moderate Resolution Imaging Spectroradiometer on Terra platform. Applied Optics, 47(36), 6796.

Meister, G., Kwiatkowska, E. J., Franz, B. A., Patt, F. S., Feldman, G. C., & McClain, C. R. (2005). Moderate-Resolution Imaging Spectroradiometer ocean color polarization correction. Applied Optics, 44(26), 5524.

Franz, B. A., Bailey, S. W., Werdell, P. J., & McClain, C. R. (2007). Sensor-independent approach to the vicarious calibration of satellite ocean color radiometry. Applied Optics, 46(22), 5068.