Soil fertility is a key determinant of sustainable agricultural productivity, particularly in semiarid regions where conventional surveys are often limited. This study evaluated the fertility status of Bardarash district, Duhok province, Iraq, by developing a soil fertility index (SFI) through an integrative framework combining geostatistical methods, machine learning (ML), GIS, remote sensing, and laboratory analyses. A total of 52 composite soil samples (0-30 cm) were analyzed for 17 physicochemical properties, including texture, bulk density, soil moisture content (SMC), organic matter (OM), soil pH, electrical conductivity (EC), cation exchange capacity (CEC), calcium carbonate (CaCO₃), and macro- and micronutrients including available nitrogen (N), phosphorus (P), potassium (K), magnesium (Mg), iron (Fe), manganese (Mn), copper (Cu), zinc (Zn), and boron (B). Principal component analysis (PCA) was applied to assign parameter weights for SFI computation, while ordinary kriging (OK) facilitated the spatial interpolation of soil properties. Additionally, ML techniques, including gradient boosting regression (GBR) and random forest (RF), were employed to enhance soil fertility prediction accuracy. The models were validated using performance metrics such as root mean square error (RMSE), mean absolute error (MAE), mean squared error (MSE), and the coefficient of determination (R²). The results indicate that GBR achieved the highest predictive accuracy, outperforming both RF and OK. Spatial analyses revealed that approximately 70% of the study area exhibits low fertility, with OM depletion, N and P deficiency, and suboptimal CEC identified as key limiting factors. In order to promote precision agriculture and sustainable land management practices in northern Iraq, these findings emphasize the significance of combining multivariate weighting approaches with geospatial ML and remote sensing to create precise, site-specific soil fertility maps.