Merging

Bio-optical Processing

Methods

The GlobColour project is the European technological solution that will fulfill the requirements for merged ocean colour products expressed by IOCCG and IOCCP.

As being a first step in the setting of forces to harmonise ocean-colour data merging (at least) in Europe, necessary preparatory steps of evaluation and characterisation shall be carried out from the beginning of the project. It is important to remind that the ocean colour data merging allows, to a given extent, to reduce the level of uncertainties of the merged quantities. However, this is only valuable when merged data are close in time and space so that there is a benefit compared to an instantaneous value. There is a trade-off analysis to conduct to make the link between the right temporal and spatial windows and the expected gain in accuracy. All these considerations lead to the fact that satellite data characterisation, both of the input data streams and the final merged products, is central to an optimal set up of merging approach.

This need for characterisation is also obviously critical for the temporal and spatial representativity of the EO measurements. So, the preparatory steps that will be carried out within GlobColour mainly consist in answering the very fundamental questions:

1. What is the level of compatibility from one sensor to the another and to the future ones?
2. What are the error bars that can be associated to each of the retrieved information from each sensor; this covers validation of the retrieved information (e.g., Chlorophyll) through comparison with in situ information and/or using the known chain of uncertainties starting from the sensor radiometric performance, accounting for the atmospheric disturbances knowledge and temporal/spatial variability up to the modelling (forward/retrieved) errors.
3. Given this present and future framework what is the best merging procedure among those already emerging and implemented internationally (binning, blended analysis, optimal interpolation, mixing of all available cross-calibrated radiance prior to a common inversion, ) ?
4. Knowing that any method will show advantages/weaknesses, what shall be the best set of criteria for selection ? In other words what shall be the trade-off to apply for the final selection?

Elements of answers will be found in the report that will be shortly issued by the dedicated IOCCG working group on "Ocean Colour Data Merging".

Number of merging techniques have been tried and are theoretically eligible to be appropriate for GlobColour (e.g. Binning / Averaging;/ Error-weighted averaging / Subjective Analysis / Blended Analysis / Optimal Interpolation / Objective Analysis / Wavelet Analysis / Machine Learning Analysis / Spectral Bio-Optical Modeling). The merging techniques selection will be done after evaluation of several of the above-mentioned methods on a pre-defined dataset and fulfillment of evaluation criteria (question 4 above).

The European Service for Ocean Colour - GlobColour is an ESA Data User Element Project