Coupling wavelet transform with multivariate adaptive regression spline for simulating suspended sediment load: Independent testing approach

Abstract

ABSTRACT Accurate prediction of suspended sediment load (SSL) of a river is very important as it directly affects the performance of the corresponding hydraulic structures. SSL can give valuable information on the catchment erodibility and deposition of the sediment being produced through scouring phenomenon. Despite the hydraulic approaches for studying the scouring/sedimentation processes in streams, hydrologic approaches may provide valuable information about SSL deposition magnitudes as well as its temporal distribution in relation with the streamflow power. A hydrologic-based approach through coupling the wavelet-based processed signals and multi adaptive regression spline (WMARS) methodology is suggested in the present paper for the first time to predict SSL values of rivers using the simultaneous streamflow and SSL records. The developed models were assessed through the most powerful k-fold testing data scanning procedure. The obtained results showed the superiority of the proposed hybrid WMARS models over the single MARS and traditional sediment rating curve techniques. Due to the presence of hysteresis effects during the study periods, adaptation of k-fold testing assessment approach is very necessary to get better insight about the models performances.