Evaluation of Statistical Regression Models in Storm-Wise Sediment yield

In this study the efficiency of statistical regression models and, also, the relationships between rainfall characteristics with runoff and sediment yield were investigated in the plot scale. For this purpose, data from runoff and sediment yield of 12 storms during October 2010 to October 2011 were used in the plot scale, which, installed in representative basin of Khamsan located at south of Kurdistan province. Hence, after each storm-runoff event, amounts of runoff and sediment were collected and measured after transporting to the laboratory. The storms data were also acquired from meteorological station located in the basin. Results showed that, sediment yield of plots have the most correlation with runoff volume, maximum intensities of storm in 60, 55, 65, 25 and 20 minute with correlation coefficient of (0.887, 0.821, 0.818, 0.814, 0.813 and 0.810). in the other hand, the runoff production of plot has the most correlation with rainfall height and 3th & 4th quartile of rainfall in significant level of 99 percent. It has been also revealed that Logarithmic model with the highest determination coefficient (0.825), efficiency coefficient (0.84) and the least value of relative error (11.57) and Root mean square (0.058) is the best model for predicting plot’s runoff production. It is Followed by power and inverse models with values of efficiency coefficient (0.8 & 0.794), determination coefficient (0.769 & 0.768) and relative error of verification (4.73 & 3.84). Results, therefore, demonstrated that multi-variation models is better to explain sediment yields of storm followed by S curve, cubic, quadratic, exponential, power and linear models.