Retrieving the evolution of vertical profiles of Chlorophyll-a from satellite observations using Hidden Markov Models and Self-Organizing Topological Maps

- We link sea-surface observations with vertical distributions of oceanic parameters.
- We reduce the dimension of the vectors associated by using Self-Organizing Maps.
- The classes generated are used by a discrete Hidden Markov Model.
- Use of the topological properties of the SOMs to improve the inversion method
- We apply the methodology to retrieve the vertical distribution of Chlorophyll-a.

We present a statistical method, denoted PROFHMM, to infer the evolution of the vertical profiles of oceanic biogeophysical variables from sea-surface data. This method makes use of discrete Hidden Markov Models whose states are defined through Self-Organizing Topological Maps. The Self-Organizing Topological Maps are used to provide the states of the Hidden Markov Model, as well as improve its parameters. After introducing the general principles of PROFHMM, we present the results obtained in a case study in which the evolution of the vertical profiles of Chlorophyll-a was inverted from sea-surface data. We applied PROFHMM for the reconstruction of the evolution of the vertical distribution of Chlorophyll-a at BATS, by training it on the numerical outputs of the NEMO-PISCES model, and reproducing the evolution of this model by using a sequence satellite observations. We obtained a root mean square error of 0.0399 ng/l for the validation year 2008.