Estimating the concentration of physico chemical parameters in hydroelectric power plant reservoir

The United Nations Educational, Scientific and Cultural Organization (UNESCO) defines the amazon region and adjacent areas, such as the Pantanal, as world heritage territories, since they possess unique flora and fauna and great biodiversity. Unfortunately, these regions have increasingly been suffering from anthropogenic impacts. One of the main anthropogenic impacts in the last decades has been the construction of hydroelectric power plants. As a result, dramatic altering of these ecosystems has been observed, including changes in water levels, decreased oxygenation and loss of downstream organic matter, with consequent intense land use and population influxes after the filling and operation of these reservoirs. This, in turn, leads to extreme loss of biodiversity in these areas, due to the large-scale deforestation. The fishing industry in place before construction of dams and reservoirs, for example, has become much more intense, attracting large populations in search of work, employment and income. Environmental monitoring is fundamental for reservoir management, and several studies around the world have been performed in order to evaluate the water quality of these ecosystems. The Brazilian Amazon, in particular, goes through well defined annual hydrological cycles, which are very importante since their study aids in monitoring anthropogenic environmental impacts and can lead to policy and decision making with regard to environmental management of this area. The water quality of amazon reservoirs is greatly influenced by this defined hydrological cycle, which, in turn, causes variations of microbiological, physical and chemical characteristics. Eutrophication, one of the main processes leading to water deterioration in lentic environments, is mostly caused by anthropogenic activities, such as the releases of industrial and domestic effluents into water bodies. Physico-chemical water parameters typically related to eutrophication are, among others, chlorophyll-a levels, transparency and total suspended solids, which can, thus, be used to assess the eutrophic state of water bodies. Usually, these parameters must be investigated by going out to the field and manually measuring water transparency with the use of a Secchi disk, and taking water samples to the laboratory in order to obtain chlorophyll-a and total suspended solid concentrations. These processes are time- consuming and require trained personnel. However, we have proposed other techniques to environmental monitoring studies which do not require fieldwork, such as remote sensing and computational intelligence. Simulations in different reservoirs were performed to determine a relationship between these physico-chemical parameters and the spectral response. Based on the in situ measurements, empirical models were established to relate the reflectance of the reservoir measured by the satellites. The images were calibrated and corrected atmospherically. Statistical analysis using error estimation was used to evaluate the most accurate methodology. The Neural Networks were trained by hydrological cycle, and were useful to estimate the physicalchemical parameters of the water from the reflectance of visible bands and NIR of satellite images, with better results for the period with few clouds in the regions analyzed. The present study shows the application of wavelet neural network to estimate water quality parameters using concentration of the water samples collected in the Amazon reservoir and Cefni reservoir, UK. Sattelite imagens from Landsats and Sentinel-2 were used to train the ANN by hydrological cycle. The trained ANNs demonstrated good results between observed and estimated after Atmospheric corrections in satellites images. The ANNs showed in the results are useful to estimate these concentrations using remote sensing and wavelet transform for image processing. Therefore, the techniques proposed and applied in the present study are noteworthy since they can aid in evaluating important physico-chemical parameters, which, in turn, allows for identification of possible anthropogenic impacts, being relevant in environmental management and policy decision-making processes. The tests results showed that the predicted values have good accurate. Improving efficiency to monitor water quality parameters and confirm the reliability and accuracy of the approaches proposed for monitoring water reservoirs. This thesis contributes to the evaluation of the accuracy of different methods in the estimation of physical-chemical parameters, from satellite images and artificial neural networks. For future work, the accuracy of the results can be improved by adding more satellite images and testing new neural networks with applications in new water reservoirs.