m5 algorithm

Suspended sediment load estimation is one of the most important and complicated debates on sediment transport and river engineering. Therefore, the variety of methods developed to estimate suspended sediment load in rivers. According to the empirical nature of these methods, the results have low accuracy and vary so widely from one method to another. Recently, with advances in computer science, various algorithms such as tree based methods have been developed. In this study, sediment rating curve method along with new algorithms such as M5 and Genetic Programing (GP) are used for estimating suspended sediment load in rivers. Flow and sediment discharge data at five hydrometric stations, Behbahan and Cham Nezam on Maroon River, Jow Kanak on Allah River and Moshrageh and Shadegan on Jarahi River are used in present study. The efficiency of tree models, in all stations, was greater than sediment rating curve method, and the RMSE performance of M5 method is 7 to 41 percent superior to sediment rating curve method. The results of this study indicate the close proximity of both M5 and GP efficiency, which according to the simple and conceptual structure of M5, this method is proposed to estimate suspended sediment load in river streams.

Snow depth is the most common parameter used for the assessment of water resources in the mountainous areas. Therefore، knowledge about spatial distribution of snow depth is the substantial knowledge of watershed characteristics. At present research، it was tried to estimate the spatial distribution of snow depth using regression kriging based on M5 model tree. Therefore، location of 216 points was selected systematically، and then snow depth was measured with a Monte - Rose sampler in Yazd-Sakhvid region. Then، 30 terrain parameters were derived from a digital elevation model using SAGA software. Our results indicated that channel network base level، stream power and wetness index were the most important parameters in decision-tree model. The correlation coefficient of 90% confirmed the strong performance of regression kriging model. Moreover، this method is very simple، so it is recommended the regression kriging model is being used to estimate spatial distribution of snow depth in other regions.