Browsing by Author "Tesfaye Kebede Gurmessa"

Now showing 1 - 1 of 1

Open Access
Numerical investigation on flow and transport characteristics to improve long-term simulation of reservoir sedimentation
(2007) Tesfaye Kebede Gurmessa; Westrich, Bernhard (Prof. Dr.-Ing. habil.)
Long-term prediction of the quantity and spatial distribution of sedimentation is required in the planning and management of reservoirs. Numerical models, conceptual models, empirical models, scale models, or a combination of them can be used in order to predict long-term sedimentation of reservoirs. Some approaches are complex while others tend to oversimplify addressing practical questions related to sediment management. This work has attempted to address simplified methodologies to predict the amount and spatial distribution of reservoir sedimentation. Two complementary approaches were specifically assessed: the numerical and the data-driven modeling approaches. Numerical modeling of long-term sedimentation processes involve model, parameter, and data uncertainties and often require very high computational costs. Model conceptualization and solution procedures involved in long-term simulation of sedimentation processes of riverine systems are challenging. The first part of the study deals with the investigation of simplification of numerical simulations. The study begins with investigations made on the basic system response of reservoirs to unsteady inputs. By making use of hypothetical domains and periodic inputs, efforts were made to try to specify the theoretical concepts. The response of the reservoirs was investigated by using aggregate parameters such as theoretical residence time, critical erosion discharge, amplitude and the frequency of inputs. The analytical solutions of ideal continuously stirred reactors and/or plug flow reactors as compared with the real two-dimensional advection-dispersion numerical solutions of the TELEMAC modeling system were evaluated. Criteria for the refinement of quasisteady steps were investigated revealing the importance of reservoir shape, range of discharges, and residence time. The work then continues by investigating into long-term simulation and simplification strategies. A thorough evaluation of hydrological, topographical, and sediment data was conducted on the Lautrach reservoir of River Iller. Using the two-dimensional depth-integrated TELEMAC modeling system, sensitivity studies were conducted with relevance to data aggregation, temporal and spatial discretization, coupling methods, turbulence models, and sediment gradation. Based on the simplifications from the preliminary studies, long-term reservoir sedimentation was calibrated and validated with fully unsteady simulation. The validated morphological simulations were tested for the extent of the applicability of quasisteady approximation considering large steady time steps. It was found out that except for extreme discharges and low flows with high sediment concentrations, the quasisteady approximation with large time steps can be successfully applied without major discrepancy from the fully unsteady simulation. Comparisons of the prediction of bed evolution were made using complete unsteady, mixed-unsteady-quasisteady, and steady approximations. In the second part of the work, an approach on the use of principal component regression in modeling the spatio-temporal bed evolution processes of riverine system was developed. The principal component analysis made on long-term bed evolution simulation indicated that only the first four principal components represent some 95 percent of the variance. This indicated that a significant simplification and representation of the spatio-temporal simulated data can be made in a condensed form. Multiple regression models were then investigated between the first four principal component scores and the flow and the sediment inputs, resulting in a very good correlation. The reconstruction of reservoir bed evolution resulted in an excellent agreement when multiple regression was used between the principal component score and the time series of discharge, change in discharge, and suspended sediment concentration. The model was also reasonably validated with acceptable uncertainty for ranges outside the period of reconstruction. It was found that the regression parameters are dynamic due to the dynamic nature of reservoir bed as well as flow and sediment parameters. The work is a step forward towards simplifying the complex and computationally demanding task of modeling of long-term reservoir sedimentation by assimilating dynamic and data modeling techniques.