The California Current merged satellite-derived 4-km dataset

Mati Kahru, mkahru@ucsd.edu                                                            Updated : May 17 30, 2020

While ocean color satellites measure more or less the same variables, systematic differences exist between the respective products of different sensors (e.g. Fig. 1A below) and between satellite products and in situ measurements. Using methods described in Kahru et al. (2012, 2015, 2018), a merged dataset of surface chlorophyll-a concentration (Chla) using data from multiple satellite sensors has been created and is regularly updated for the California Current region. The method uses match-ups of each sensor with a large dataset of in situ measurements (e.g. over 12,000 for surface chlorophyll measurements) to minimize the differences between each sensor and in situ measurements AND between the overlapping satellite sensors. Modern satellite Chla retrievals start with OCTS (Nov., 1996) and are followed by SeaWiFS (1997-2010), MODIST (2000-present), MERIS (2002-2012), MODISA (2002-present), VIIRS-SNNP (2012-present), OLCI-A (2017-present), VIIRS-JPSS1 (2017-present), OLCI-B (2018-present), SGLI (2018-present). Daily remote sensing reflectance (Rrs) data at 4 km resolution (9 km for OCTS and SeaWiFS, 1 km for OLCI-A, OLCI-B, SGLI) are used to calculate daily products for each sensor using the regionally optimized algorithms, after which they are merged. Note that this is different from the merged products at 1-km resolution available at http://www.wimsoft.com/CAL/.  The 1-km products are merged between the standard products of all available sensors without applying the optimized algorithms to the Rrs.  

Visible and infrared satellite data are limited by the presence of clouds. In order to reduce the amount of missing data due to clouds, 5-day running composites are created, i.e. the pixel value for each day is an average of the valid values for 5 consequtive days (from 2 days earlier to 2 days later). Those 5-day running means of daily datasets are then composited into 5-day composites. The remaining missing data are interpolated between the previous and the next 5-day product. The same procedure is applied again to produce 2x interpolated data. Fig. 2 shows the effects of temporal interpolation. Monthly composites are also made from the interpolated 5-day products.

Fig. 1A shows the systematic bias that exists between VIIRS-SNPP and MODISA standard Chla values in the beginning of 2020. Fig. 1B shows the difference between the optimized MODISA Chla (MODISA2015) and the standard MODISA Chla.  In February, 2019 the whole time series was reprocessed by blending with the NASA OCI .method at low Chla  The MODISA2015 algorithm produces higher values at in situ Chla > 1 mg m-3 but is practically the same at lower Chla. The standard NASA Chla algorithm uses ocean color index (OCI) at low Chla levels (Hu et al. 2012). As our in situ dataset has few very low values, we adopted the NASA blending approach which makes more reliable estimates at low Chla. We blend our Calfit2015 estimates (Kahru et al. 2015) with OCI estimates at low Chla. As data from new sensors  such as  VIIRS-JPSS1 (2017-present), OLCI-A (2016-present), OLCI-B (2018-present) and SGLI (2018-present) become available, we gradually include the data from these new sensors after evaluating the differences.

Net Primary Production (NPP) estimates are calculated from the 5-day merged Chla, merged daily PAR (from MODISA, MODIST, VIIRS-SNNP, VIIRS-JPSS1) and daily SST-OI data (Reynolds et al. 2007) using a modified VGPM (Behrenfeld and Falkowski 1997) called VGPM-CAL (Kahru et al. 2009). By using daily PAR and daily SST and with 5-day Chla we are assuming that Chla is changing relatively slowly and we can generate daily NPP products (Kahru et al. 2016).

Export Flux of Carbon (EF, mg C m-2 day-1) and export fraction (EF/NPP) are calculated according to the regionally fitted algorithm folowing Kelly et al. 2018. This approach provides subtantial improvement compared to the previously used global algorithms (Laws et al. 2011). In May, 2019 the algorithm has been revised to use different coefficients (EF = 0.284 * NPP + 9.75). This algorithm is based on the in situ dataset of M. Stukel (Stukel York Type II regression) and has been described in Kahru et al. 2020.

Fig 1

 

Fig. 1. Pixel-wise scatter of (A) VIIRS-SNPP standard Chla versus MODISA standard Chla and (B) optimized MODISA Chla (MODISA2015) versus standard MODISA Chla. Both datasets use same day pixels in 2020 (first 81 days). 81 days).

 

Fig. 2. An example of a 5-day Chla product of January 1-5, 2003 (left), the interpolated 5-day (middle) and the 2x interpolated (right) 5-day product.

 

The datasets are in Zipped files and can be downloaded using the hyperlinks below. The latest major update of all products except SST was completed on 26-March-2019,

Chla Daily

Chla Daily

Chla Daily

Other

1996

1997

1998

Chla 5-day all years

1999

2000

2001

Chla 5-day interpolated

2002

2003

2004

Chla 5-day interpolated 2x

2005

2006

2007

Chla monthly ;            Chla monthly PNG files

2008

2009

2010

 

2011

2012

2013

SST-OI 5-day HDF4 files ; SST-OI 5-day PNG files

2014

2015

2016

SST-OI monthly HDF4 ; SST-OI monthly PNG files

2017

2018

2019

 
2020      

NPP Daily

NPP Daily

NPP Daily

Other

1996

1997

1998

Net Primary Production 5-day, HDF4

1999

2000

2001

 

2002

2003

2004

Net Primary Production monthly

2005

2006

2007

 

2008

2009

2010

 

2011

2012

2013

 

2014

2015

2016

 
2017 2018 2019
2020      

EF, EFr Daily

EF, EFr Daily

EF, EFr Daily

Other

1996, 1996

1997, 1997

1998, 1998

EF = export flux of Carbon, same units as NPP

1999, 1999

2000, 2000

2001, 2001

EFr = export fraction, i.e. EF/NPP

2002, 2002

2003, 2003

2004, 2004

EF and EFr are calculated using daily NPP

2005, 2005

2006, 2006

2007, 2007

EF 5-day all years  ; EFr 5-day all years

2008, 2008

2009, 2009

2010, 2010

 

2011, 2011

2012, 2012

2013, 2013

EF Monthly all years ; EFr Monthly all years

2014, 2014

2015, 2015

2016, 2016

 
2017, 2017 2018, 2018 2019, 2019
2020, 2020      

 

The datasets are on a grid of 540 (width) x 417 (height) with approximately 4000 m step in HDF4 format.  The upper-left corner (lat, lon) is 45N; -140E; the lower-right corner is 30.03597N; -115.5454E. Chla values in each pixel of unsigned byte are log10-scaled and can be calculated from the pixel value (PV) as: Chl (mg m-3) = 10^(0.015 * PV - 2.0), i.e. 10 to the power of 0.015 * PV - 2.0. Pixel values 0 and 1 (black in Fig. 1) and 255 (white in Fig. 1) are considered invalid and must be excluded from any statistics.  PV = 1 is used for coastline and has to be excluded too. The annotation (color bar) is written into the dataset. When reading with Matlab the unsigned byte variable is sometimes reported as signed byte (int8, values from -127 to 128) and not as unsigned byte (values from 0 to 255) and values over 128 become negative. A simple fix is to add 256 if the signed pixel value is negative. Net Primary Production (NPP) was calculated according to the modified VGPM algorithm (Kahru et al. 2009). Pixel values are signed 2-byte integers in mg C m-2 day-1 with values -32767 or 32767 meaning no data.

The SST data are converted from the global AVHRR OI dataset (http://podaac.jpl.nasa.gov/dataset/NCDC-L4LRblend-GLOB-AVHRR_OI ) and use the same scaling as in the full resolution dataset (http://www.wimsoft.com/CAL/Readme.htm).

References

Behrenfeld, M.J., P.G. Falkowski (1997), Photosynthetic rates derived from satellite based chlorophyll concentration. Limnol. Oceanogr. 42, 1-20. https://doi.org/10.4319/lo.1997.42.1.0001

Hu, C., Z. Lee, and B. Franz (2012), Chlorophyll a algorithms for oligotrophic oceans: A novel approach based on three-band reflectance difference, J. Geophys. Res., 117, C01011, https://doi.org/10.1029/2011JC007395

Kahru, M., Kudela, R., Manzano-Sarabia, M., Mitchell, B.G., 2009. Trends in primary production in the California Current detected with satellite data. J. Geophys. Res. Ocean. 114, 1-7. https://doi.org/10.1029/2008JC004979

Kahru, M., R.M. Kudela, M. Manzano-Sarabia and B. G. Mitchell (2012), Trends in the surface chlorophyll of the California Current: Merging data from multiple ocean color satellites, Deep-Sea Research II, 77-80, 89-98, http://dx.doi.org/10.1016/j.dsr2.2012.04.007.

Kahru, M., R.M. Kudela, C.R. Anderson, B.G. Mitchell (2015), Optimized merger of ocean chlorophyll algorithms of MODIS-Aqua and VIIRS. IEEE Geoscience and Remote Sensing Letters, 12, 11, https://doi.org/10.1109/LGRS.2015.2470250

Kahru, M., Z. Lee, B.G. Mitchell and C.D. Nevison (2016), Effects of sea ice cover on satellite-detected primary production in the Arctic Ocean, Biology Letters, 12: 20160223. http://dx.doi.org/10.1098/rsbl.2016.0223.

Kahru, M., M.G. Jacox and M.D. Ohman (2018), CCE1: Decrease in the frequency of oceanic fronts and surface chlorophyll concentration in the California Current System during the 2014-2016 northeast Pacific warm anomalies, Deep-Sea Research I, https://doi.org/10.1016/j.dsr.2018.08.007

Laws, E.A., D'Sa, E., Naik, P. (2011), Simple equations to estimate ratios of new or export production to total production from satellite-derived estimates of sea surface temperature and primary production. Limnol. Oceanogr. Methods 9, 593-601. https://doi.org/10.4319/lom.2011.9.593

Reynolds, R.W., T.M. Smith, C. Liu, D.B. Chelton K.S. Casey, and G. Schlax (2007), Daily high-resolution blended analyses for sea surface temperature. J. Climate, 20, 5473-5496, https://doi.org/10.1175/2007JCLI1824.1