Estimation of exchangeable sodium percentage from sodium adsorption ratio of salt-affected soils using traditional and dilution extracts, saturation percentage, electrical conductivity, and generalized regression neural networks

摘要

Soil sodicity is best evaluated by the exchangeable sodium percentage (ESP); however, the determination of this index is laborious and time consuming. Alternatively, the sodium adsorption ratio (SAR) is a simpler index that is commonly used to estimate soil sodicity. The objective of this research is to estimate ESP using four approaches: (1) SAR of saturated paste (SAR(e)), and SAR of 1:5 extracts (SAR(1:5)), (2) a conversion factor (CF) as a function of saturation percentage (theta(SP)), (3) electrical conductivity of 1:5 extracts (EC1:5), and (4) Generalized Regression Neural Networks (GRNN). Approximately 120 surface soil samples were collected from the Jordan Valley region and ESP, SAR(e), SAR(1:5), (theta(SP)), soil texture, and soil hydraulic conductivity (HC) were determined. The GRNN model (i.e., Approach 4) gave the most accurate estimates for the ESP and was able to handle the heteroscedasticity of the data. Meanwhile the traditional dilution extracts (1) showed that soil ESP was highly related to SAR(e) and to SAR(1:5); the CF-theta(SP) approach (2) gave better estimates for prediction of ESP. Moreover, EC1:5 (3) gave reasonably accurate estimation of ESP and could be used as a screening test for assessment of sodicity problems. For the case study site investigations, a reduction of 20% in soil HC was observed when SAR(e) increased from 0 to 3.5 or ESP increased from 0 to 6, indicating that this reduction occurred at ECe < 3 dS m(-1) for all soils. While the theta(SP) approach reduced the effect of heteroscedasticity of the data on the predictive model ability, the GRNN models can accurately predict the ESP based on easy-to-obtain soil features. Our models represent a rapid and accurate estimator of soil sodicity, and therefore offer a potentially valuable tool in managing soil landscapes that are vulnerable to degradation.