Evaluation of gamma test, cluster analysis, discriminant function analysis and andrews curves methods to separate homogeneous watersheds for regional analysis of suspended sediment
Subject Areas : Farm water management with the aim of improving irrigation management indicators
Hossein
Kheirfam
1
(Ph.D Student, Department of Watershed Management Engineering, Faculty of Natural Resources, Tarbiat Modares University, Noor, Mazandaran, Iran)
Mehdi
Vafakhah
2
(Associate Professor, Department of Watershed Management Engineering, Faculty of Natural Resources, Tarbiat Modares University, Noor, Mazandaran, Iran)
Keywords: homogeneous watersheds, sediment modeling, sediment rating curve, suspended sediment estimation,
Abstract :
Sediment yield resulting from soil erosion in the watersheds is the major limitation in achieving the sustainable development and major threat to ecosystems. Therefore estimation of output sediment from watersheds is very important. Extent of watershed and deficiency of sediment measuring stations have caused us to use different indirect methods to estimate sediment, such as the use of models provided in similar watersheds. In this study 42 sediment measuring stations in south and southeast of the Caspian Sea with over 20 year period were chosen. By relating suspended sediment load to stream discharge in the times of flood , daily suspended sediment was estimated by using the daily discharge and the average of annual sediment was calculated. By using the Gamma Test, 13 factors affecting sediment yield were reduced to 5 main factors and by using cluster analysis, discriminate function analysis and andrews curves, study stations were put in homogeneous groups. For each homogeneous group obtained from any one of the mentioned homogenization methods and by using five main factors selected, regression models were developed to estimate the average of annual suspended sediment. Error rates and accuracy of prepared models by using statistical indices of RE, RBIAS and RRMSE were calculated according to observed data. Results indicated that all three homogenous techniques had better results than those of the general model and Andrews Curves with 38.12 and 45.91% RE, 53.16 and 33.11% RRMSE and -0.01 and 0.01 RBIAS in calibration and validation stages, respectively and had better performance than those of two methods i.e. Cluster Analysis and Discriminate Function Analysis for homogenizing of south and southeaster Caspian Sea watersheds based on sediment yield. Also peak discharge (Qp) has the most impact on the average of annual suspended sediment changes.
