The Relationship Between Physiognomic Characteristics of Tamarix aphylla and Seidlitzia rosmarinus with Morphometric Parameters of Khour Va Biabanak County Nebkhas Using Regression Methods and Artificial Neural Network

Document Type : Research Paper


1 ardakan university

2 Assistant Professor, Department of Nature Engineering, Faculty of Agriculture &Natural Resources, Ardakan University.IRAN

3 Assistant Professor, Department of Nature Engineering, Faculty of Agriculture & Natural Resources, Ardakan University, P.O. Box 184, Ardakan, Iran.


Nebkha and the plant that creates it together form a system. nebkha formation parameters have a significant effect on the transfer rate of aeolian deposits. Adequate information on nebkha will help plan more accurately and efficiently and better manage wind erosion-prone lands to identify appropriate wind erosion programs and deal with this phenomenon. This study aimed to compare regression methods and artificial neural networks to investigate the relationship between the quantitative characteristics of Tamarix and Seidlitzia Rosmarinus plant species and quantitative parameters of nebkha. The regression methods used in this study include PCR, PLS, and OLS. In this study, the plant characteristics used are plant height, length, width, and type, and morphometric characteristics include nebkha length, height, slope, and width. The number of sampling points was 80 selected randomly from nebkha in Khour va Biabanak County. 70% of the data was used for training the network and 30% for validation. According to the results of this study, the highest R2 between nebkha length and Seidlitzia Rosmarinus plant characteristics is related to the OLS method with 0.8, followed by nebkha area and width. R2 levels in the neural network were lower, i.e., 0.76. In the case of the Tamarix species, the highest R2, i.e., 0.797, is related to the characteristics of the plant with nebkha length, followed by nebkha area and width. This value is 0.78 in the neural network method. Moreover, the evaluation results of different predictive models showed that the OLS model is superior to other models.