Determination of Soil Properties Affecting Soil Aggregate Tensile Strength in a Semiarid Region of Iran Using a Hybrid Algorithm

作     者：Hojjatnooghi, Fatemeh Shirani, Hossein Pazira, Ebrahim Besalatpour, Ali Asghar Torkashvand, Ali Mohammadi 

作者机构：Islamic Azad Univ Dept Soil Sci Sci & Res Branch Tehran Iran Vali E Asr Univ Rafsanjan Coll Agr Dept Soil Sci Rafsanjan Iran 

出 版 物：《COMMUNICATIONS IN SOIL SCIENCE AND PLANT ANALYSIS》 (土壤科学与作物分析通讯)

年 卷 期：2021年第52卷第17期

页      面：1981-1992页

核心收录：

学科分类：0710[理学-生物学] 09[农学] 0903[农学-农业资源与环境] 0901[农学-作物学] 0703[理学-化学] 

基　　金：Science and Research Branch, Islamic Azad University, Tehran, Iran Islamic Azad University 

主　　题：Soil aggregate stability Brazilian method PSO algorithm ICA algorithm SVR algorithm 

摘      要：The present study was conducted to determine soil properties affecting soil aggregate stability in southeast Iran. For this purpose, 169 soil samples were collected at 0-20 cm depth, and some soil properties such as soil bulk density (BD), electrical conductivity (EC), calcium carbonate equivalent (CCE), and soil organic matter (OM) were determined. Subsequently, soil aggregate stability (as quantified by tensile strength (TS)) were determined using the indirect (Brazilian) method. Furthermore, soil properties affecting soil aggregate stability were determined using the Particle Swarm Optimization-Imperialist Competitive Algorithm-Support Vector Regression (PSO-ICA-SVR) hybrid algorithm, and after the sensitivity analysis, the importance of each selected property in terms of its impact on soil aggregate stability was recognized. The results indicated that OM, BD, EC, and dg (geometric mean particle diameter), in general, influence soil aggregate stability, and this selection was carefully made by the hybrid algorithm. Beside, after modeling using the features selected by the SVR method and performing the sensitivity analysis, BD and OM, among the selected properties, had the most significant effects on soil aggregate stability. In this study, the SVR method achieved greater accuracy in predicting the TS index compared to the Multiple-Linear Regression (MLR). The RMSE value (Root Mean Square Error) for the TS index using the MLR model was 30.77, while a lower RMSE value of 10.536 was observed for the SVR model.