Prediction of infragravity wave swash motions on natural beaches using model trees

Occurrence of extreme storm waves predisposes natural beaches to erosion. During occurrence of these storm waves, infragravity swash energy plays an important role in the amount of erosion created. The available formula for infragravity swash is only based on wave heights and wave length and its accuracy is low. Recently, model trees have been introduced as one of the new methods in the data mining approaches that give all possible relations between the involved parameters. In this paper, a new model based on hydrodynamic parameters including the surf similarity, momentum flux and non-dimensional characteristic velocity parameters is presented by using the M5′ and MARS model trees. To generate and evaluate models, totally of 579 field data available in literature is used. The results indicated that the developed MARS model improves the RMSE and R values by 42% and 26%, respectively, and the M5′ model improves these values by 16% and 12%, respectively, in respect to the most common empirical model. According to sensitivity analysis of MARS model and also results of M5′ model, the non-dimensional characteristic velocity parameter showed the most correlation and the surf similarity parameter showed the least correlation with infragravity swash motions. The performance of developed models is also compared with a numerical study implemented on Tizimin beaches in Mexico.