New study helps NEODAAS pick the best algorithm for estimating chlorophyll from satellite data

News

19 May 2021

A comprehensive assessment of the performance of chlorophyll-a (Chl-a) products from current ocean colour sensors over the Atlantic Ocean has recently been published in Remote Sensing of the Environment, including contributions from NEODAAS ocean colour expert Silvia Pardo.
 

There are multiple satellite sensors which provide estimates of chlorophyll concentration, with different algorithms available for each. When supplying data to users, for example in support of research cruises, it is important for NEODAAS to understand how well different sensors and algorithms perform.

The goal of the paper was to evaluate the accuracy of the Sentinel-3A and 3B OLCI chlorophyll products for three different processors: the standard EUMETSAT 2016-2020 processing baseline PB2(1), processing baseline PB3(2), and POLYMER(3). The PB2 processor computes Chl-a using a four-band, blue-green reflectance ratio algorithm, OC4Me. The recently developed PB3 processor includes improvements such as updated system vicarious calibration gains, bright pixel correction, cloud masking, as well as the implementation of a colour index (CI) algorithm for low Chl-a waters. The POLYMER processor derives Chl-a from in-water reflectance spectral matching using 2-dimensional optimization techniques.

Figure 1: Scatter plots of AMT26, 27, 28 in situ versus S-3A OLCI Chl- a for independent (a-c) and coincident (d-f) match-ups.

Figure 1: Scatter plots of AMT26, 27, 28 in situ versus S-3A OLCI Chl-a for independent (a-c) and coincident (d-f) match-ups.


As shown in figure 1 above, POLYMER provided the best performance in the retrieval of Chl-a in both independent and coincident match-up analysis. While PB2 and PB3 performed similarly for chlorophyll values above 0.1 mg m-3 range, the performance of the colour index-based algorithm implemented in PB3 was superior to the OC4Me band-ratio algorithm of PB2 for chlorophyll concentrations below 0.1 mg m-3 range. The PB3 processor improved the overall performance, with better linear regression fit, and lower bias and RPD metrics. The effect was more evident at low concentrations, where band-ratio algorithms are more sensitive to errors in the atmospheric correction.

Figure 2: Scatter plots of AMT26, 27, 28 in situ Chl a versus MODIS-Aqua and Suomi-VIIRS
Figure 2: Scatter plots of AMT26, 27, 28 in situ Chl a versus MODIS-Aqua and Suomi-VIIRS


This work makes use of the Atlantic Meridional Transect (AMT) programme, designed to undertake sampling at high spatial and temporal resolution across a wide range of ecosystems in the Atlantic Ocean, for the validation of ocean-colour products. The quality and volume of the in situ datasets generated by AMT campaigns have increased exponentially in the past ten years. Particularly for ocean colour, this is due to the development of techniques to measure Chl-a continuously from underway optical systems, and the introduction of Fiducial Reference Measurements (FRM) protocols for the acquisition of radiometry measurements.

Figure 3 - (a) AMT26 (September–October 2016), (b.) AMT27 (September–October 2017) and (c.) AMT28 (September–October 2018) campaign tracks superimposed on Sentinel-3A OLCI Chl a composites for each campaign. The locations of OLCI-A PB2 matchups are shown in orange.
Figure 3 - (a) AMT26 (September–October 2016), (b.) AMT27 (September–October 2017) and (c.) AMT28 (September–October 2018) campaign tracks superimposed on Sentinel-3A OLCI Chl a composites for each campaign. The locations of OLCI-A PB2 matchups are shown in orange.

Silvia Pardo, NEODAAS Research Scientist, commented: "Our capacity to perform studies like this is limited by the number of in situ measurements coincident with satellite data. For early AMT cruises, there were 2-3 in situ chl-a measurements a day, so you would need to wait ten years to accrue enough data to write a paper like this! Nowadays we have 1-2 chl-a measurements per minute and they are high quality, so we can be very strict in selecting the best ones to match with the satellite data."
 
Data from the MODIS-Aqua and Suomi-VIIRS instruments were also evaluated against OLCI-3A. The analysis showed that MODIS performed similarly to the POLYMER product, while the VIIRS results were comparable to those obtained for PB3. Finally, the OLCI sensors abroad Sentinel3-A and Sentinel3-B were compared during the tandem phase coincident with AMT28, when the two satellites were flying together along the same orbit. The comparison highlighted small differences between the two sensors, with OLCI-3B being slightly more accurate than its Sentinel3-A counterpart.

The analysis used the archive of satellite date being downloaded as part of NEODAAS and utilised custom software developed by NEODAAS for data processing. 
 

Further Information

Tilstone, G.H., Pardo, S., Dall'Olmo, G., Brewin, R.J.W., Nencioli, F., Dessailly, D., Kwiatkowska, E., Casal, T. and Donlon, C. (2020). Performance of ocean colour chlorophyll a algorithms for Sentinel-3 1 OLCI, MODIS-Aqua and Suomi-VIIRS in open-ocean waters of the Atlantic. Remote Sensing of Environment, 260, art.112444. doi:10.1016/j.rse.2021.112444

References:
  1. EUMETSAT. (2018). Sentinel-3 OLCI marine user handbook. EUM/OPS-SEN3/MAN/17/907205. https://earth.esa.int/eogateway/documents/20142/1564943/Sentinel-3-OLCI-Marine-User-Handbook.pdf.
  2. Sentinel-3 OLCI L2 report for baseline collection OL_L2M_003 (2021) EUMETSAT Report, EUM/RSP/REP/21/1211386.
  3. F. Steinmetz, P.Y. Deschamps, D. Ramon (2011). Atmospheric correction in presence of sun glint: application to MERIS, Opt. Express, 19, 9783-9800, 10.1364/OE.19.009783.
  4. I. Cazzaniga, E. Kwiatkowska  (2018). Sentinel-3 OLCI Chlorophyll Index switch for low-chlorophyll waters Algorithm Theoretical Basis Document. EUMETSAT Report, EUM/RSP/DOC/18/1028360.

Other News articles:

Displaying results 11-12 (of 22)
 |<  <  2 - 3 - 4 - 5 - 6 - 7 - 8 - 9 - 10 - 11  >  >|