Excessive or deficient levels of soil micronutrients severely restrict crop production, particularly in complex semiarid drylands where spatial variability is driven by intense environmental and anthropogenic factors. Because traditional large-scale sampling is costly and time-consuming, this study aimed to predict the spatial distribution and uncertainty of soil nutrients (S, Fe, Mn, Zn, and Cu) across a 476-ha semiarid area in Thoothukudi, Tamil Nadu. A multi-sensor Digital Soil Mapping (DSM) framework driven by the Quantile Regression Forest (QRF) algorithm was employed, integrating 112 soil samples with 42 environmental covariates (topographic, multispectral, and radar). The findings revealed widespread, severe deficiencies in S (85%) and Zn (43%). The QRF model handled the extreme spatial heterogeneity effectively, yielding moderate explanatory power (R2 = 0.24 to 0.49) and relatively low prediction errors (RMSE = 1.10 to 3.70). Crucially, unlike traditional models, QRF successfully quantified spatial uncertainty; Prediction Interval Coverage Probability (PICP) values ranged from 89.0% to 94.3%. While this indicated a slight overestimation for Cu (94.3%), the models overall closely aligned with the nominal 90% confidence level and ensuring robust, risk-aware estimates. Furthermore, variable importance analysis revealed distinct pedogenic drivers: physical topographic covariates governed the stable distribution of Fe and Mn, whereas biological and anthropogenic spectral proxies (e.g., NDVI, NDBI) dictated Zn and Cu variability. Ultimately, this study demonstrates that integrating multi-sensor QRF models provides a highly scalable template for transitioning from generalized fertilization to risk-aware, Site-Specific Nutrient Management (SSNM), offering a vital decision-support tool for sustainable agricultural intensification in data-scarce dryland regions.