Appendix

Ocean color data processing

1.  Original CHL algorithm

Siswanto et al. (2011) developed YOC CHL algorithm for turbid water,

,

,

where oS is SeaWiFS Reprocessing 5.1 (R2005.1) dataset.

For non-turbid water, following standard algorithms were used;

SeaWiFS OC4V4 (O’Riley et al. 2000)

,

,

and,

MODIS OC3M (O’Riley et al. 2000)

,

,

For switching from non-turbid to turbid water, non-turbid water algorithm, turbid water algorithm and linear combination of those were used for nLw(555) 1.5, nLw(555) 2.5 and 1.5 ≤ nLw(555) ≤ 2.5 , respectively (Yamaguchi et al. 2013).

2.  Conversion for SeaWiFS Reprocessing 2010.0 (R2010.0) dataset

In order to convert the original turbid water algorithm developed for SeaWiFS R2005.1 (eqns. (1) and (2)) to the one for SeaWiFS R2010.0 dataset, linear regression analyses between SeaWiFS R2005.1 and R2010.0 Global Area Coverage (GAC) Level 2 datasets were conducted. Pixel-wise band ratios [Rrs(412)/Rrs(490) and Rrs(443)/Rrs(555)] and Rrs(555) were selected for the regression analyses, and 120 daily data for each month from 1998 to 2007 was used. Data with more valid pixels were selected to represent each month. Rrs of SeaWiFS R2005.1 was converted by nLw multiplying terrestrial solar irradiance (provided in the meta-data of SeaWiFS Level 2 file). The extraterrestrial solar irradiances (F0) for SeaWiFS wavelength are followings; F0(412)=171.18, F0(443)=188.75, F0(490)=193.37, and F0(555)=183.76.

The results of the regressions are followings;

,

,

,

where nS is SeaWiFS R2010.0 dataset.

The original CHL algorithm for turbid water (eqns. (1) and (2)) was then converted to eqns. (6) and (7) by eqns. (3) and (4). nLw(555) switching criteria (Yamaguchi et al. 2013) was also converted with eqn (5).

,

3. Conversion for MODIS dataset

A similar regression analyses were also made between SeaWiFS R2010.0 Merged Local Area Coverage (MLAC) Level 2 band ratios [Rrs(412)/Rrs(490) and Rrs(443)/Rrs(555)] and MODIS Reprocessing 2009.1/2010.0 (R2009.1/R2010.0) LAC Level 2 band ratios [Rrs(412)/Rrs(488) and Rrs(443)/Rrs(547)] with 30 daily datasets from 2002 to 2004;

,

,

,

where nM is MODIS R2009.1/R2010.0 dataset.

The converted SeaWiFS CHL algorithm for turbid water (eqs. (6) and (7)) was then converted to eqns. (11) and (12) by eqns. (8) and (9). nLw(555) switching criteria (Yamaguchi et al. 2013) was also converted with eqn. (10).

,