Spectral Response Function Concerns

There are several issues concerning the spectral response functions for the MODIS instruments, including:

  • Detector-specific or Detector-averaged?
  • Terra vs. Aqua - are they really different?
  • Out-of-band (OOB) - can the measurements be trusted?

Detector-specific vs. Detector-averaged

The following is a simple analysis showing the detector specific spectral response functions for each of the visible bands convolved with a hyper-spectral water-leaving radiance typical of clear, open ocean water.

  Band 8 - 412nm (View Graph) Band 9 - 443nm (View Graph)
Detector Aqua
Lw412nm (mW/m2/um/sr)
Terra
Lw412nm (mW/m2/um/sr)
Aqua
Lw 443nm (mW/m2/um/sr)
Terra
Lw 443nm (mW/m2/um/sr)
1 1.53831 1.56724 1.38255 1.38223
2 1.53831 1.56621 1.38255 1.38237
3 1.53834 1.56578 1.38242 1.38241
4 1.53842 1.56551 1.38239 1.38240
5 1.53850 1.56505 1.38236 1.38243
6 1.53871 1.56486 1.38234 1.38249
7 1.53901 1.56414 1.38232 1.38254
8 1.53930 1.56341 1.38232 1.38262
9 1.53994 1.56308 1.38232 1.38275
10 1.53998 1.56308 1.38249 1.38275
AVG 1.53888 1.56484 1.38240 1.38250
% Diff 100*(T-A/A) 1.68663 100*(T-A/A) 0.00698490
  Band 10 - 488nm (View Graph) Band 11 - 531nm (View Graph)
Detector Aqua
Lw 488nm (mW/m2/um/sr)
Terra
Lw 488nm (mW/m2/um/sr)
Aqua
Lw 531nm (mW/m2/um/sr)
Terra
Lw 531nm (mW/m2/um/sr)
1 1.03870 1.04801 0.399058 0.400915
2 1.03870 1.04796 0.399058 0.400915
3 1.03870 1.04790 0.399058 0.400927
4 1.03868 1.04798 0.399090 0.400930
5 1.03869 1.04795 0.399110 0.400930
6 1.03872 1.04797 0.399124 0.400926
7 1.03874 1.04791 0.399128 0.400902
8 1.03877 1.04790 0.399136 0.400603
9 1.03882 1.04790 0.399107 0.400603
10 1.03889 1.04790 0.399107 0.400603
AVG 1.03874 1.04794 0.399098 0.400825
% Diff 100*(T-A/A) 0.885213 100*(T-A/A) 0.432887
  Band 12 - 551nm (View Graph) Band 13 - 667nm (View Graph)
Detector Aqua
Lw 551nm (mW/m2/um/sr)
Terra
Lw 551nm (mW/m2/um/sr)
Aqua
Lw 667nm (mW/m2/um/sr)
Terra
Lw 667nm (mW/m2/um/sr)
1 0.314212 0.316653 0.0209927 0.0209237
2 0.314212 0.316646 0.0209927 0.0209271
3 0.314220 0.316653 0.0209922 0.0209275
4 0.314228 0.316668 0.0209941 0.0209261
5 0.314223 0.316657 0.0209943 0.0209266
6 0.314226 0.316650 0.0209943 0.0209274
7 0.314232 0.316650 0.0209934 0.0209249
8 0.314225 0.316651 0.0209944 0.0209235
9 0.314213 0.316617 0.0209950 0.0209232
10 0.314257 0.316617 0.0209942 0.0209232
AVG 0.314225 0.316646 0.0209937 0.0209253
% Diff 100*(T-A/A) 0.770560 100*(T-A/A) -0.325838
  Band 14 - 678nm (View Graph)
Detector Aqua
Lw 678nm (mW/m2/um/sr)
Terra
Lw 678nm (mW/m2/um/sr)
1 0.0184740 0.0184604
2 0.0184740 0.0184601
3 0.0184737 0.0184603
4 0.0184738 0.0184604
5 0.0184737 0.0184610
6 0.0184737 0.0184613
7 0.0184741 0.0184604
8 0.0184741 0.0184611
9 0.0184737 0.0184628
10 0.0184727 0.0184628
AVG 0.0184738 0.0184610
% Diff 100*(T-A/A) -0.0688040

The differences between the detectors are insignificant. There appears to be no need for detector-specific spectral response functions. A single, average sepectral response function will suffice. In fact, all bands have several detectors with identical spectral response functions. The reason for this is discussed below under "Measurement Issue".

Terra vs. Aqua

The above, detector-specific analysis shows that the differences between sensors are slightly greater than the inter-detector differences, but it may be that this is solely due to the OOB component, which is often quite similar between the two sensors (see the associated figures). The OOB component does differ, but may be more of a magnitude or measurement issue than any real difference.

Measurement Issue

The MODIS RSRs were measured by SBRS with a Spectral Measurement Assembly (SpMA). In the SpMA, a monochromator is used to produce light of a single wavelength at an exit slit. The slit was measured by MODIS, and the output wavelengths of the SpMA were varied. Unfortunately, for one output wavelength of the SpMA, there is a small spectral variation along the slit, e.g. the upper region of the slit has a slightly different wavelength than the center. The MODIS detectors measured different parts of the slit, which resulted in shifts of the center wavelengths of the measured RSRs of the detectors relative to each other. SBRS determined that the central area of the slit contains the correct information, and shifted the RSRs of the outer detectors such that they matched the center wavelength of the central detectors. (In some cases, some of the outer detectors did not produce valid measurements, and their RSRs were copied from nearby detectors.) Therefore, the corrected RSRs of all detectors have the same center wavelength. The amount of the shift is less than 0.1% (e.g. at 500nm, a shift of less than 0.5nm). The magnitude and direction of the shift is consistent with an independent characterization of the SpMA (without the use of MODIS). However, it is unclear whether using the corrected RSRs achieves superior results than simply using a common RSR for all detectors of a band.

The detector-averaged center wavelength (in-band) for the Terra and Aqua ocean bands are:

Band Aqua(nm) Terra(nm) Difference(nm)
8 412.463 411.896 0.567
9 442.220 442.114 0.106
10 487.428 486.995 0.433
11 530.110 529.722 0.388
12 547.198 546.954 0.244
13 665.997 665.757 0.240
14 677.580 676.973 0.607
15 746.831 746.612 0.219
16 866.877 866.348 0.529

The difference between the center wavelengths for Aqua and Terra are within the magnitude of the detector specific adjustments made to compensate for the SpMA instrument artifact. It is unclear as to whether or not the differences between Aqua and Terra are real or an artifact of the measurement, or a combination.

Out-of-Band Component

There has been some concern raised regarding the OOB component of the spectral response functions for both MODIS instruments. MCST does not have confidence in the OOB component for the spectral response functions, and in fact do not use them. Experience with SeaWiFS has shown that the out-of-band component can be important, especially for ocean processing, so it was decided to use the OOB component for MODIS processing as well in order to maintain consistency.

It has been suggested that the OOB for MODIS is too high. The effect of the OOB component is evident in the bandpass averaged water absorption coefficients, especially in the blue bands (412 and 443nm). This topic was addressed on the Ocean Color Forum

If the OOB component for MODIS is in error, there will be an effect on the ocean color processing. This effect will vary depending upon the source function to which the spectral response function is applied (e.g. Rayleigh, water absorption, etc.).