NIR Correction

Chlorophyll-based NIR correction ramp-in

The near-IR correction used in the level-1 to -2 processing currently makes slight changes the chlorophyll and water-leaving radiance values in regions of low chlorophyll. This feature of the current implementation runs contrary to the assumption that the near-ir water-leaving radiance is zero at low chlorophyll, and also complicates the vicarious calibration. A modification to the algorithm that removes these changes has been implemented.

NIR iteration control

The iteration control was also modified to improve the behaviour. The following changes have been implemented:

  • damping between iterations (averaging of NIR reflectance between iterations)
  • for Siegel, iteration stops when chl changes by 2% (previously 20%).
  • for Arnone, iteration stops when average NIR water-leaving reflectance changes by less than 2%, or Rrs670 < 0.0.
  • seed chl changed from 0.3 to 0.0. seed Rrs670 for Arnone set to 0.0.
  • iteration reinitialization on any iteration where chl retrieval fails, using:
    • chl = iter*5.0
    • Rrs670 = 5.0*(0.00032 + 0.00021*chl)
      previously, if chl retrieval failed, the iteration was reinitialized with a seed chl of 5, but only on the first iteration.

Update to IR absorption coefficients

The current implementation of the NIR correction uses the absorption due to water in its formulation. The values used are derived from published values for the nominal center wavelengths of SeaWiFS NIR bands. Given the strong absorption of water in the NIR and the 40 nm band passes for the SeaWiFS NIR bands, these numbers are underestimated. A revised set of bandpass-normalized numbers were generated and implemented, based on the same published values for water absorption cited in Siegel et al. (2000).

Revised backscatter model for NIR bands

The current operational NIR correction uses Loisel and Morel (1998) to estimate particulate backscattering at 555nm and Morel (1988) to extrapolate this value into the NIR. While the initial impetus for the Siegel et. al. correction was to apply a NIR correction in highly productive case 1 waters, in practical application it is applied most often in turbid or case 2 waters. The Loisel and Morel model is explicitly case 1, and not applicable to case 2 waters. Gould et. al., (1999) developed a spectral dependence model for the scattering coefficient in case 1 and case 2 waters. This model was implemented in place of the Morel (1988) model as being more appropriate to the practical application of the NIR correction in SeaWiFS.

A reflectance-based estimate of backscatter at a reference wavelength has also been adopted. This reflectance-based backscatter estimate was proposed by Arnone and Stumpf (1998). It uses reflectance at 670nm.

Since the are no in situ data in the NIR with which the correction can be validated, the angstrom product has been chosen as a proxy. The basis for this lies in the assumption that the aerosol properties over NIR-corrected waters should be similar to those over adjacent waters where the correction has not been applied. Below are two sets of images from two HRPT scenes. One set for each scene shows the NIR correction as implemented in the operational processing (with the addition of the vanishing NIR), and one shows the reflectance based backscatter estimate and Gould et. al., spectral dependence model. Notice that the operational version shows artifacts of the chlorophyll product influencing the angstrom product. The reflectance based version shows much fewer, and smaller, artifacts. (Rrs_670 is shown in place of chlorophyll here as Rrs_670 was used to estimate backscatter in the visible.) The chlorophyll product for the reflectance based correction and the true color images for each scene are shown for reference.

Chlorophyll-based (operational) Reflectance-based (proposed)
Chlorophyll Rrs_670 Chlorophyll
Angstrom_510 True Color
Chlorophyll-based (operational) Reflectance-based (proposed)
Chlorophyll Rrs_670 Chlorophyll
Angstrom_510 True Color