An Artificial Intelligence Approach to the Assessment and Prediction of Soil Quality Dynamics

Abstract

The adverse effects of climate change, including land misuse, improper agricultural practices, and global warming, have a detrimental impact on soil health, fertility, productivity and quality. The degradation of soil, a fundamental component of the ecological system, poses a significant threat to the viability of sustainable land use practices, thereby impeding the rational and effective utilization of resources. Consequently, in order to ensure the sustainability of agricultural practices, it is essential to consider the reliability of soil quality determination methods and their suitability for large-scale implementation. The objective of this study was to predict soil quality using only the basic properties of soil (sand, clay, silt, organic matter, pH, electrical conductivity, lime, nitrogen, phosphorus, potassium) with artificial neural networks (ANN), one of the artificial intelligence algorithms that have attracted attention in recent years. The soil quality index (SQI) values of the soils within the Lake Van basin, which is characterized by a continental climate, were found to range between 0.381 and 0.703. Furthermore, the correlation coefficients (R) obtained between the actual data and the predicted data during the training, validation, and testing phases of the soil quality prediction with ANN were found to be 0.83, 0.83, and 0.71, respectively. The spatial distribution pattern of the actual and predicted values obtained in the SQI maps created using the Kriging-Simple-Spherical model, one of the geostatistical methods, in the study area, was found to be similar. The study demonstrated that incorporating additional soil properties into the model is essential for achieving more precise results.