APPLICATAION OF A NEURAL NETWORK TO MONITOR STREAM SURFACE WATER QUALITY USING SATELLITE REMOTE SENSING DATA

摘要

The investigation of earth's resources is an important aspect of remote monitoring application. Through the data received, land use and surface covering and environmental quality in rivers, dams, and lakes can be detected. This research incorporates multivariate regression, artificial neural network, and discriminate analysis to examine and compare the relationships of optical spectrum and water quality. The investigation will determine the possibility of using the images obtained through automatic remote monitoring to determine the changes in river water quality and its index derived from dissolved oxygen, biological oxygen demand, suspended solid, and ammonia nitrogen. Concurrent in situ surface water quality measurements, optical (SPOT) data were obtained in selected locations in five different days. Although significant correlations were not observed between optical data and water quality parameters, the result of this investigation has shown, based on the water predictions from monitoring stations that the analysis from multivariate regression is not as good as the results obtained from artificial neural network in the study area. Likewise, a neural network algorithm is applied to estimate the transfer functions between the major characteristics of surface water index and the satellite optical data. The results also show that the estimation accuracy, with acceptable level for major characteristics of surface water index using the neural network pairing up with the conjugate gradient decent searching method is more feasible than those from discriminate analysis. The results also indicate that conjugate gradient decent searching algorithm can assist to improve the categorization. However, this result of limited optical data learning algorithm needs to be further confirmed by more case studies. The technique still needs to be refined in detail in order to detect differences within the typical range of these water quality index found in the area under study. Basically, it is evident that artificial neural network has the potential and feasibility of monitoring water qualities and index.