Modeling of Suspended Sediment Yield in Ardabil Province watersheds using PCA and Multiple Regression Analysis

Modelling the SSY in Ardabil Province watersheds using Principal Component Analysis (PCA) and Multiple Regression Analysis (MRA) was the aim of the present study. The available suspended sediment data of 29 watersheds were used to develop the SSY estimation models through validation by six watersheds. The 15 different physiographic, climatic, geomorphic, geological variables were used in the modeling after applying PCA. Finaly, the MRA models were evaluated based on the percentage of error, and Adj-R2 as statistical measures. The results of PCA showed that the first four principal component explain about 86.39% of the total variance, which were selected as the main components for MRA input variables. The most influencial variables includes main river length, maximum precipitation, maximum elevation, and area. According to the results the Best Model with maximum height and maximum precipitation variables, Forward and Stepwise methods with perimeter variable, the Backward method with minimum elevation and maximum elevation, area, perimeter, total main stream length, maximum precipitation, mean precipitation, annual precipitation, and sensitive geological formations were selected. According to the Backward model, Minimum elevation (0.073), maximum elevation (0.001), area (0.002), perimeter (0.053), total river length (0.01), maximum precipitation (0.001), mean precipitation (0.008), and erosive geologic formations (0.082) were the final significant parameters in estimation of SSY. The drainage density and vegetation cover were excluded from the modelling process and the selected variables had a great impact on SSY over the study area, and the results can be generalized to the ungauged watersheds to SSY estimation.