Classification of cultivated soils in Al-Kharj, Saudi Arabia based on natural radionuclides using artificial neural networks

Abstract

A great awareness of safe crop production is needed by selecting the suitable uncontaminated soil, and good seed types, etc. In this study, artificial neural network (ANN) was used to classify thirteen cultivated soils, in adjacent regions, in Al-Kharj Governorate, Saudi Arabia based on their natural radionuclide contents detected using gamma-ray spectrometer and geographical coordinates and altitude measured using GPS system. A total of 3497 patterns were collected. Specific activities (238-U, 40-K, 137-Cs and 232-Th in Bq/kg) were acted as inputs to an ANN classifier. In addition, geographical coordinates and altitude acted as inputs to the ANN classifier. The best predictive power for the classification of soils from the thirteen sites was achieved using 140 hidden neurons in the hidden layer of the ANN classifier. Most of soil data not included in the training data was correctly classified with an overall classification rate of 94.63%. The obtained results indicate that natural radionuclide contents detected by gamma-ray spectrometer and geographical coordinates and altitude measured by GPS in combination with ANN classifier is a viable tool for soil classification. The outcome of this research may have many benefits to environmental, soil and crop specialists in Saudi Arabia. Key words: Soil, natural radionuclides, Saudi Arabia, artificial neural network, classification.