Akademik Veri Yönetim Sistemi
INTERNATIONAL JOURNAL OF ENVIRONMENTAL SCIENCE AND TECHNOLOGY, cilt.23, sa.2, 2025 (SCI-Expanded, Scopus)
The quality of agricultural soil has become a subject of considerable academic interest due to a number of factors. These include increased awareness of the importance of sustainable agriculture, reduction of agricultural land due to urbanization and population growth, and harmful environmental consequences of these developments. In this context, a new approach using fuzzy logic and artificial intelligence to assess soil quality, which is important for environmental sustainability, was evaluated. The objective of this study is twofold: first, to utilise the Hesitant Fuzzy Analytic Hierarchy Process (AHP) and Standard Scoring Function (SSF) approaches to ascertain the Soil Quality Index (SQI) of agricultural lands in the Samsun Province situated in the Central Black Sea Region of Turkiye; and second, to evaluate the accuracy of models such as Artificial Neural Networks (ANN) and XGBoost in predicting the obtained SQI values. Additionally, the obtained soil quality index scores were verified by comparison with biomass data for the area obtained using NDVI. To this end, both linear and nonlinear methods were employed. It is evident from the results that the estimation of soil quality indices using linear and nonlinear methods was facilitated by ANN and the XGBoost algorithm. XGBoost exhibited the lowest error rate when estimating the soil quality index using both linear and nonlinear methods. The values of R2, RMSE, and MAE for the estimation of linear and nonlinear SQI values using the XGBoost algorithm were 0.95, 0.01, and 0.08 for the linear model and 0.97, 0.03, and 0.02 for the nonlinear model, respectively. The predictive capability of the ANN algorithm was comparable to that of XGBoost, with both methods exhibiting high predictive accuracy. In multiple-comparison tests of the datasets, the observed and predicted values obtained by linear and nonlinear methods differed significantly (p < 0.01). High correlations were observed between NDVI values obtained from Sentinel-2 satellite images and the SQI. These similarities were also consistent with the spatial distribution maps obtained by geostatistical analysis. The study concluded that soil quality could be evaluated using artificial intelligence and remote sensing techniques, and these approaches should be validated in more comprehensive studies. The study concluded that soil quality could be evaluated using artificial intelligence and remote sensing techniques, and these approaches should be validated in more comprehensive studies.