The improvement of wavelet-based multilinear regression for suspended sediment load modeling by considering the physiographic characteristics of the watershed

Abstract

The aim of this study is to model a relationship between the amount of the suspended sediment load by considering the physiographic characteristics of the Lake Urmia watershed. For this purpose, the information from different stations was used to develop the sediment estimation models. Ten physiographic characteristics were used as input parameters in the simulation process. The M5 model tree was used to select the most important features. The results showed that the four factors of annual discharge, average annual rainfall, form factor and the average elevation of the watershed were the most important parameters, and the multilinear regression models were created based on these factors. Furthermore, it was concluded that the annual discharge was the most influential parameter. Then, the stations were divided into two homogeneous classes based on the selected features. To improve the efficiency of the M5 model, the non-stationary rainfall and runoff signals were decomposed into sub-signals by the wavelet transform (WT). By this technique, the available trends of the main raw signals were eliminated. Finally, the models were developed by multilinear regressions. The model using all four factors had the best performance (DC = 0.93, RMSE = 0.03, ME = 0.05 and RE = 0.15).

Fire protection priorities in the oak forests of Iran with an emphasis on vertebrate habitat preservation

“…To model the distribution, the MaxEnt algorithm approach was utilized using R version 4.2.1 (R Development Core Team). Model performance was evaluated using the Area Under the receiver operating Curve (AUC = ROC) and the True Skill Statistic (TSS), as recommended by Allouche, Tsoar 69 ) and Zipkin, Grant 70 , 71 .…”