Hybridisation of fuzzy systems and genetic algorithms for water quality characterisation using remote sensing data

We propose the characterisation of a coastal water scheme using a combination of remote sensing data and in situ measurements. The characterisation is performed using most discriminate water-mapped parameters, such as concentration of suspended particulate matter 'SPM', suspended chlorophyll 'Chl', water turbidity 'Turb' and water transparency 'SDD'. The relation between these parameters and the multispectral reflectance is non-linear. Hybridisation of fuzzy model and genetic algorithm (GA) using remote sensing data allows the estimation of these parameters. The principle of our estimation is mainly based on a set of fuzzy rules extracted automatically from the training data in a two-step procedure. First, fuzzy rules are generated using an unsupervised fuzzy clustering of the input data. Second, a GA is applied to tune these rules. Our contribution consists in providing global and partial optimisation rules, in which a proposed chromosome structure adapted to our database achieves a correlation coefficient close to 98.42. In addition to that, we propose a valuable tool to evaluate the pollution degree through estimated indices maps. From these maps, a pollution signature draw is constructed and examples from specific sites are compared to highlight their pollution degree, then the most polluted case is retrieved. This approach is tested in two sites having different pollution properties, namely ETM+ of Landsat 7 and MODIS data covering, respectively, Algiers and Manche sea.