This research introduces the use of virtual sensors to estimate the concentration of physico-chemical parameters and metals in monitoring water quality of reservoirs Amazon through artificial neural networks (ANN) and images of remote sensing. A factor analysis of the variables considered in the study confirmed the relationship of the first factor with Secchi disk, Total Iron, PO4, Total P, TSS and Turbidity on a single factor, as these have a high reflectance and good energy absorption by satellite sensors. These elements were determined by ANN's, producing satisfactory results approach 100% between observed and estimated. The tests resulted in a good approximation, the first band Secchi disk depth, total Fe, STS, and turbidity of the water reservoir. In the specific case of the parameters PO4 and Total P, besides the problem of the small number of sampling stations available data and the variability inherent in the hydrological cycle of the region, it was found, through the interpretation of images, lack of similarities between the data used in training and validation of RNA. Overall, the study demonstrated the effectiveness of the application of virtual sensors in monitoring water quality of reservoirs in the Amazon by satellite imagery, providing a precise and less expensive alternative resources in the process of environmental monitoring.
|
