Application of proximal sensing approach to predict cation exchange capacity of calcareous soils using linear and nonlinear data mining algorithms

• Published in: Journal of soils and sediments Vol. 24; no. 6; pp. 2248 - 2267
• Main Authors: Karami, Ali, Moosavi, Ali Akbar, Pourghasemi, Hamid Reza, Ronaghi, Abdolmajid, Ghasemi-Fasaei, Reza, Lado, Marcos
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.06.2024; Springer Nature B.V
• Subjects: Acetates; Acetic acid; Algorithms; Arid zones; Calcareous soils; Cation exchange; cation exchange capacity; Cation exchanging; Cations; cost effectiveness; Data analysis; Data mining; Data processing; Earth and Environmental Science; Environment; Environmental Physics; Estimation; Exchange capacity; Infrared analysis; Iran; Land use; Least squares method; prediction; Prediction models; Radial basis function; Reflectance; Regression; regression analysis; Sec 1 • Soil Organic Matter Dynamics and Nutrient Cycling • Research Article; Semi arid areas; Semiarid zones; Sodium; Sodium acetate; Soil chemistry; Soil fertility; Soil properties; Soil Science & Conservation; Soils; Spectral reflectance; Spectroscopy; Support vector machines; Wavelengths; Iran; Spectrotransfer function; Stepwise multiple linear regression; Support vector regression; Soil spectroscopy; Multiblock analysis; Partial least squares regression; Pedotransfer functions
• ISSN: 1439-0108; 1614-7480
• DOI: 10.1007/s11368-024-03825-7

Purpose The cation exchange capacity (CEC) is a pivotal soil attribute that influences soil chemistry, fertility, and productivity. Nevertheless, the conventional techniques employed for CEC measurements present challenges in terms of complexity, cost, and laboriousness. Hence, there is a demand for expedited, cost-effective, streamlined alternative methodologies that can yield accurate outcomes. The objective of this study was to employ and compare various techniques, including Pedotransfer Functions (PTFs) based on fundamental soil properties, support vector regression (SVR), sequential orthogonalized partial least squares (SOPLS) as a multiblock data analysis method, and Spectrotransfer Function (SPTF) utilizing visible-near infrared (VNIR) and mid infrared (MIR) diagnostic wavelengths to estimate the CEC of calcareous soils with diverse land uses in the semi-arid region of Fars province, Iran. Materials and methods A total of 130 samples were gathered from the soils of the study region, CEC was measured using sodium acetate, and the spectral reflectance in the VNIR and transmission in the MIR regions were measured, and prediction models were created using linear support vector regression (L-SVR), radial basis function support vector regression (RBF-SVR), partial least squares regression (PLSR), and multiblock data analysis algorithms, after different spectral preprocessing methods. Results and discussion The results generally indicated that spectroscopy models performed better than PTFs in predicting CEC with the multiblock SOPLS showing the best results (R 2  = 0.92, RMSE = 1.67 cmol (+) kg −1 , and RPIQ = 4.34). The performance of the models followed the order: SOPLS > SPTF > L-SVR > RBF-SVR. Conclusion Our findings indicate that spectroscopy coupled with SOPLS analysis can be a robust, viable, fast, cheap, and efficient alternative assessment method with acceptable accuracy for estimating soil CEC in calcareous soils, instead of the difficult, costly, and cumbersome conventional measurement approaches or other estimation methods.