Processing and analysis of remotely sensed images for water quality monitoring based on computational intellegence

Harmful Algal Blooms (HABs) are serious problems for global aquaticecosystems and they appear to be increasing in the surface waters aroundthe world. Chlorophyll-a concentration is an important index for evaluatingwater quality. Remote sensing of ocean color provides a convenient methodof determining the concentrations from spectral information. In the openocean, it is not difficult to derive empirical algorithms relating the receivedradiances to surface concentrations of Chlorophyll-a. In some waters,however, this is much more difficult due to the presence of highconcentrations of suspended sediments and dissolved organic material,which dominate the spectral signal of Chlorophyll-a. Thus, water qualityretrieval combining in-situ measurements with remote sensing images is adifficult task as well as challenge. It is here that the field ofhydroinformatics can contribute. Firstly, remote sensing images need to bepre-processed to get correct spectral information of Chlorophyll-a aselaborated in this thesis. Secondly, since artificial neural networks (ANNs)have proven to be successful in modelling a variety of geophysical transferfunctions, a neural network approach is employed here to model the transferfunction between the in-situ Chlorophyll-a concentrations and the satellitereceiveddata. By inputting the whole remote sensing image into the model,Chlorophyll-a concentrations for the whole area could be obtained. It wasfound that by adding classification according to suspended sediment andprior knowledge on the underlying processes to the data, it was possible tomodel the transfer function to a much higher accuracy than the routinelinear regression models or other non-linear regression models. In thisresearch, a neural network using the three visible Landsat Thematic Mapperbands, classification indices according to suspended sediment andknowledge-based indices of the study area were used as inputs. In this way,the classification of Chlorophyll-a can be used to understand where theeutrophic prone areas are likely to be found