Potential of ANN modeling in development of Sediment Rating Curves for a Large Trans-Himalayan Alluvial River

Código

I-EBHE0215

Autores

Archana Sarkar

Tema

WG 2.6: Water systems analysis for integrated planning & management

Resumo

The assessment of the volume of sediments being transported by a river is required in a wide spectrum of water resources problems such as the design of reservoirs and dams; hydroelectric power generation and water supply; transport of sediment and pollutants in rivers, lakes and estuaries; determination of the effects of watershed management; and environmental impact assessment. Therefore, accurate estimate of sediment transported by a river is of vital importance. Rivers emerging out from the Himalayan region transport the sediment at a very high rate. Himalayan and Tibetan region cover only about 5% of the Earth's land surface, but they supply about 25% of the dissolved load to the world oceans. In order to estimate the sediment load in rivers, field engineers widely use the rating curve approach based on statistical regression analysis. However, the complexity of the problem cannot be addressed by simple regression and curve-fitting techniques. An intrinsic problem of rating curve technique is the high degree of scatter, which may be reduced but not eliminated. The sediment concentration for a flood on a rising discharge differs from that on the falling discharge under some conditions like flatter gradients and constricted channels, especially for the alluvial Himalayan Rivers. This phenomenon results in looped sediment-discharge curves known as hysteresis. A number of researchers have proposed different methods for the improvement of rating curve estimations by employing various statistical correction factors, using nonlinear regression, or classifying the sediment data into different ranges. The problem of sediment rating becomes even more complex for a river like the Brahmaputra, because of its Himalayan origin, steep slopes, alluvial nature and large spatio-temporal variations in various hydro-meteorological variables as well as landuse/landcover including high seismic activity in the region. Therefore, the artificial neural network (ANN) models which are capable of modelling the complex non-linear hydrological processes with limited data are the best choice for the present modelling problem. The present paper presents development of sediment-discharge rating relation using artificial neural networks (ANN) technique as well as regression based conventional Sediment rating curve (SRC) methods for two important gauging sites of the Trans-Himalayan mighty Brahmaputra River within India. Comparison of the two methods has also been made. The sediment rating curves have been developed using fifteen years of daily data of discharge and Suspended sediment concentration. The performance of the models has been evaluated using various statistical performance indices as well as sedimentographs and scatter plots. ANN models were found to be significantly superior than the conventional SRCs, the latter are being frequently used by field engineers. The study suggests that for both gauging sites, i.e., on the main Brahmaputra River and on the northern tributary namely, Subansiri River, the ANN modelling approach is able to capture the inherent nonlinearity in the sediment-discharge rating relationships. The pooled average relative overestimation and underestimation errors (PARE) were less for ANN models producing very accurate estimation of the entire spectrum of suspended sediment concentration for both gauging sites, a very important aspect considering the braided nature of river.