EJSSEurasian Journal of Soil ScienceEurasian J Soil Sci2147 - 4249Federation of Eurasian Soil Science Societieshttp://ejss.fesss.org/10.18393/ejss.190180010.18393/ejss.1901800Evaluation of soil quality in a cornell-based polye ecosystem in a Karstic Area: Integration of geostatistics and machine learningHüseyin Alper Aktop Sevda Altunbaş Beritanli saltunbas@akdeniz.edu.trBayram Çağdaş Demirel 152291-3012025-10-212026-02-232026-03-032026-04-01 Copyright © 2016 The authors and Federation of Eurasian Soil Science Societies Assessing soil quality and establishing processes to control quality is critically important in environmentally sensitive karst systems. This study aimed to analyze the spatial behavior of soil quality and the dominant soil properties controlling it in the karstic polje system of Gembos Plain (Southern Turkey) using the Cornell Soil Quality Index (SQI). SQI values were calculated using physical and chemical data obtained from 72 soil samples (0–30 cm). According to analysis results SQI values ranged from 53 to 59, with an average value of approximately 56. Semivariogram analysis showed that SQI exhibited moderate spatial dependence; the Gaussian variogram model was selected as the most appropriate model due to lowest error values and RMSSE close to 1. The spatial distribution obtained by ordinary kriging revealed that low-to-medium quality classes were dominant across the area and that geochemical micro-scale patterns prevailed over distinct morphological zoning. In addition, Random Forest and XGBoost models showed the highest performance in machine learning analyses (R² = 0.74 and 0.71). Variable importance analysis has shown that soil reaction (pH) is the primary control factor determining SQI variability, while P, Fe, and K are effective at a secondary level. Also soil organic matter plays a secondary role in explaining SQI variability due to its narrow range of variation. The results reveal that soil quality in karstic polje conditions is controlled by chemical constraints rather than physical properties, and that the combined use of geostatistical modeling and machine learning significantly improves the interpretability of soil quality analyses.Soil quality, Karstic polje, Cornell Soil Quality Index, Geostatistical modeling, Machine learning.