Somayeh Mafi’s scientific contributions

One of the most vital tasks for coastal engineers is to calculate the gross longshore sediment transport rate (LSTR) to control the shoreline erosion and beach evolution. In the past decades, several empirical formulas or parametric models have been proposed for predicting the gross LSTR as a function of the breaking wave characteristics, bed materials and beach slope. The main downside with these formulas is that they give wide range of different rate of predictions; consequently their reliability under the changing hydrodynamic conditions is very uncertain. In the present study, an alternative approach based on the Regression Trees (M5') was applied to present a new formula for prediction of LSTR in terms of the longshore sediment transport coefficient (c). Several high-quality data sets were employed, which comprise of the wave parameters, sediment characteristics and the longshore transport rate. Based on the obtained results, the Shields parameter was selected as the main input variable, while the longshore sediment transport coefficient (c) was given as the output parameter. The M5' algorithm was employed for building and evaluating the regression trees, which showed that regression trees can be used successfully for the prediction of LSTR. Finally, the performance of the new M5'-based formula in the prediction of LSTR was compared with the previous formulas. The results indicated that the error statistics of the regression trees were less and it is evidently predicts the LSTR more accurate than the previous mentioned formulas.