Estimation of the Exchangeable Sodium Percentage from the Sodium Adsorption Ratio for salt-affected soils in Bolivia

Abstract

<p>In order to obtain a more cost-time efficient way to determine the sodicity of salt-affected soils, this study aims to generate a regression model to predict the Exchangeable Sodium Percentage (ESP) from the Sodium Adsorption Ratio (SAR<sub>e</sub>). Based on a database with 84 soil samples from the High Valley of Cochabamba (Bolivia), two linear models were generated: <em>ESP= 0.9725 SAR + 1.5766 </em>(R<sup>2</sup>=0.85, RSE=4.47) and <em>ESP= 0.9197 SAR + 0.3813 </em>(R<sup>2</sup>=0.71, RSE=0.91)<em> </em>with square root transformation. Subsequently, through a set of 18 testing samples and a T-test of paired samples between the predicted ESP and measured ESP values, the efficiency of the generated models was verified with a value of p= 0.063 and 0.209, respectively, in contrast to p= 0.285 from the US Salinity Lab referential model. To improve the performance of the generated models, could be necessary a stratification according to soil sodicity levels and additional samples for the calibration dataset.</p>