Spatial prediction of saline and sodic soils in rice‒shrimp farming land by using integrated artificial neural network/regression model and kriging

ABSTRACT

In the context of widespread saline and sodic soil, mapping and monitoring spatial distribution of soil salinity and sodicity are important for utilization and management in agriculture lands. In this study, two-stage assessment was proposed to predict spatial distribution of saline and sodic soils. First, artificial neural network (ANN) and multiple linear regressions (MLR) model were used to predict sodium adsorption ratio (SAR) and exchangeable sodium percentage (ESP) based on soil electrical conductivity (EC) and pH. Then, the Kriging interpolation method combined with overlay mapping technique was used to perform saline spatial predictions in the study area. The model accuracy level is evaluated based on coefficient of determination (R²) and root mean square error (RMSE). In the first stage, the values of R² and RMSE of SAR and ESP were 0.94, 0.17 and 0.94, 0.24 for ANN, and 0.35, 0.52 and 0.34, 0.76 for MLR, respectively. Similarly, in the second stage, the RMSE of ANN-Kriging were much closer to 0 and relatively lower than MLR-Kriging and Kriging. The results show that ANN-Kriging can be used to improve the accuracy of mapping and monitoring spatial distribution of saline and sodic soil in areas that develop the rice-shrimp cultivation model.

KEYWORDS:

• Exchangeable sodium percentage
• sodium absorption ratio
• salt-affected soils
• spatial analysis prediction
• rice–shrimp cultivation

Acknowledgements

The authors are grateful to Prof. N.N. Hung from the Can Tho University for his comments that are helpful to the improvement of the manuscript. We thank the editors and the reviewers for their helpful comments, suggestions and revisions on this manuscript.

Disclosure statement

No potential conflict of interest was reported by the authors.