Multi-station streamflow forecasting using wavelet denoising and artificial intelligence models

In this research, the hybrid of threshold based wavelet denoising with Extreme Learning Machine (ELM) and Least Square Support Vector Machine (LSSVM) models would be investigated in order to forecast Snoqualmie watershed daily Multi-Station (MS) streamflow. For this purpose, firstly, the watershed outflow was forecasted using models of ELM and LSSVM only with one station individually without any pre-processing. So, MS-ELM and MS-LSSVM were applied for using all sub-basins data synchronously. Ultimately, the sub-basins streamflow were denoised using wavelet based thresholding approach, and next, in a MS framework, the purified signals were imposed into the ELM and LSSVM models. It was obtained the ELM preference compared to the LSSVM, and MS model with compared to the individual sub-basin model, considering denoised data with respect to the noisy data, for example, DCLSSVM=0.82, DCELM=0.84, DCMS-ELM=0.90, DCdenoised-MS-ELM=0.93.