NAOC @ Aston

NAOC or Neural network Algorithms for Ocean Colour, is an European Union Framework V funded program, which seeks to improve the retrieval of Ocean Colour products using neural networks and associated technologies.

General

The problem is conceptually simple. Polar orbiting satellites measure the radiance (basically reflected sunlight) at the top of the atmosphere. Over the ocean the intensity of light received in different wave-bands of the visible and near infra-red regions can tell us something about the underlying atmosphere and the colour of the ocean surface (to about 10 meters depending on conditions). The measurement of ocean colour is useful because we can assess chlorophyll-A concentrations from space and thus infer ocean primary productivity and infer something about the global carbon cycle.

Retrieval of Ocean Colour

Although the problem seems simple enough, a major complication arises due to the atmospheric contribution to the ocean colour signal. The atmosphere is not clear in the visible and infra-red regions of the spectrum. Two effects dominate - Rayleigh scattering at a molecular level occurs predictably as a function of wavelength and can be corrected by the use of information in the near infra-red region of the spectrum, at least to first order. Particle absorption and scattering is more difficult to correct for, so regions with large dust contents - such as of the west of the Sahara are difficult to handle. The colour of the underlying water is also a factor - ocean colour experts distinguish between case I water (clear and deep) and case II waters which are turbid and generally shallower.

NAOC will try and improve retrieval of ocean colour in all conditions, and also attempt a retrieval of individual pigments.

Aston Role

Within the Neural Computing Research Group at Aston we have a great deal of experience of applying neural networks to remote sensing problems and the development of novel principled neural network related technologies. In NAOC we will seek to exploit some recently developed methods such at the Generative Topographic Mapping and Bayesian methods for retrieval of Ocean Colour.

We will work across many themes in NAOC:

• in Work Package 1 we will explore the application of unsupervised learning methods to the characterisation of the aerosol and water type.
• in Work Package 3 we will investigate the use of direct models to retrieve pigment concentrations from water leaving radiances.
• in Work Package 4 we will investigate the use of spatial, Bayesian methods to jointly retrieve Chlorophyll concentrations, and aerosol parameters in a principled probabilistic manner.
• in Work Package 5 we will help develop the complete chain prototype.

Last modified: 18/02/2002