Application of empirical, regression and artificial intelligence methods for the sediment transport in natural streams of the Aegean region

The suspended sediment load can be determined by means of several methods such as direct measurements at the sediment gauging stations, sediment rating curve, regression methods, artificial intelligence methods and empirical methods which are based on experimental works. Although direct measurement is the most reliable method, it is very expensive, time consuming, and, in many instances, problematic for inaccessible sections, especially during floods. Because of this, measurements of suspended sediment load are carried out in longer periods compared to precipitation and flow measurements.In this study the suspended sediment load of Gediz, Küçük Menderes and Büyük Menderes Rivers which are the main water resources of Eagen Region, are investigated. In the light of composed scenerios, sediment rating curve, regression methods, artificial intelligence methods and empirical methods are tested for the four stations in the region. In the first part it is aimed to establish suspended sediment load models by means of observations. For this purpose, multi-linear regression, multi-nonlinear regression, artificial neural networks, and adaptive neural inference fuzzy system models are developed. The study results have revealed that sediment rating curve method which is frequently preferred in Turkey has accuracy limitations but artificial intelligence methods in general have better performance for all the stations.In the second part, it is aimed to establish suspended sediment load models related to river and catchments characteristics by means of empirical approaches. Some of the parameters which are needed for the empirical approaches are obtained by land and laboratory work. While Brooks Method is determined as an appropriate approach for the Eagen Region through the empirical approaches, the calibration of this method for the Eagen Region is accomplished by the Genetic Algorithm.In the third part, daily suspended sediment loads from daily precipitation and flow data are predicted. At the present situation, the suspended load measurements are carried once a month and then the sediment rating curves are obtained from these measurements. Using these curves, daily suspended loads are predicted from daily flow rates. As pointed out earlier, however, sediment rating curves do not provide satisfactory results. Hence, this study has successfully employed artificial neural networks to predict missing suspended sediment load data for Acısu Station on Gediz river.At the end of the study, the regional analysis is carried out. The analysis has involved the together treatment of upstream stations, donstream stations, and all the four stations. Also, in this section, a sensitivity analysis is performed. The results have revealed that the transport of the suspended sediment load is very sensitive to the particle diameter.