Ion-adsorption rare-earth deposits supply over 90% of the global market’s heavy rare-earth elements (HREEs). The genesis of these deposits, particularly HREE deposits, has garnered significant attention. To elucidate the metallogenic mechanisms of HREE deposits, a comprehensive study of the weathering profile of granite was conducted in Jiangxi Province, South China. This study focuses on the following two aspects: the petrogeochemistry of HREE-rich granite and the enrichment and fractionation of rare-earth elements (REEs) during the weathering process. The results suggest that the Dabu granites are a typical peraluminous, high-K, calc-alkaline granite series with high silica content (SiO₂: 74.5%–76.4%), relatively low phosphorus content (P₂O₅: <0.05%), and high HREE content (ΣLREE/ΣHREE: 0.16–0.66). Weathering advances the decomposition of minerals and the release of elements. REEs are mainly fixed in the regolith by scavengers, mainly clays, Fe–Mn oxides, and carbonates, and ΣREE can reach 799 ppm in the B horizon. However, HREEs tend to migrate further and preferentially combine with Fe–Mn oxides and carbonates as compared to LREEs, leading to a significant fractionation of REEs in the regolith (ΣLREE/ΣHREE = 0.2–1.1). Additionally, the differential weathering of REE-bearing minerals and the precipitation of secondary REE-bearing minerals are also vital for REE fractionation.

Soil C:N:P stoichiometry reflects biological inputs, land management, and mineral weathering, but how these controls vary with depth remains unclear in tropical basaltic landscapes. We analyzed 500 soil profiles sampled at 0–20, 80–100, and 180–200 cm in northern Chengmai County, Hainan Island, to test directional predictions about weathering–carbon associations, surface enrichment, land-use contrasts, and nutrient covariation. Mean Chemical Index of Alteration (CIA) was high at all depths (92.8–93.7), while molar SiO₂/Al₂O₃ decreased from 14.0 to 10.5. Surface organic carbon (Corg) was positively associated with CIA, Al₂O₃, and TFe₂O₃. From surface to deep soil, Corg, total nitrogen (TN), and total phosphorus (TP) decreased by 68.4%, 64.4%, and 36.3%, respectively. Within-profile enrichment factors were greater for Corg and TN than TP (Holm-adjusted p < 0.001), accompanying larger declines in C:P and N:P than C:N. Fruit orchards, dryland, and paddy fields had higher mean surface TP and lower C:P and N:P than grassland and forest land. Corg differences among land uses disappeared at 180–200 cm. Corg–TP and TN–TP associations weakened with depth, whereas TP remained positively associated with Fe- and Al-rich components. These results support unequal biological and mineral influences on nutrient depth patterns.

Heavy metal transfer from agricultural soils to edible crops is an important concern for food safety and sustainable land management in tropical regions, where intensive weathering and heterogeneous parent materials can decouple total soil metal concentrations from plant uptake. In this study, 1120 soil samples and 1093 crop samples representing 55 crop species and five operational crop groups were collected across Hainan Island, China. Geological background and 38 soil physicochemical properties, including available micronutrients such as B, Cu, Fe, Mn, Mo, S, and Zn, were integrated into Random Forest models to evaluate their associations with the bioaccumulation factors (BCFs) of Pb, Cr, Cd, As, and Hg. Mean BCFs on a dry-weight basis were 1.144 for Cd, 0.111 for Hg, 0.057 for Cr, 0.046 for As, and 0.009 for Pb. Cd transfer was relatively higher in vegetables, whereas woody fruits generally showed lower transfer. Five-block spatial validation with a 5 km training buffer yielded BCF R² values of 0.045 for Cd, 0.122 for Cr, 0.060 for Hg, −0.079 for Pb, and −0.061 for As, which were substantially lower than the corresponding OOB estimates. Removing soil Cr reduced the spatial R² for Cr BCF from 0.122 to 0.002, whereas removal of soil Cd had little effect on Cd BCF performance. Excluding crop group also reduced predictive performance for Cd, Cr, and Hg. Geological background was associated with total soil metal distributions, whereas soil pH, Fe/Al-related variables, exchangeable cations, and crop group showed associations with variation in soil-to-crop transfer. Accumulated local effects analysis showed nonlinear fitted relationships between BCF and several key soil properties within the observed data range. Overall, total soil metal concentrations alone provided an incomplete indication of crop accumulation. These results highlight the value of considering soil acidity, exchangeable base-cation status, geological background, and crop type together when evaluating heavy-metal transfer in highly weathered tropical agricultural soils.

Stream-sediment anomalies around the Sawayardun gold deposit integrate mineralized detritus with variable lithological and hydraulic backgrounds. This study evaluates whether multielement coherence and spatial continuity can distinguish Au-related signals from that mixed background in 1080 primary samples and 35 field duplicates. Au is strongly right-skewed (median 1.62 ng/g, P95 6.55 ng/g, maximum 754.3 ng/g). Correlation and hierarchical clustering identify Au–As–Sb–Mo as the most persistent mineralization-related association, with a secondary Bi contribution; Ag and Cu provide important local support, whereas Co–Cr–Ni–Sn–W mainly record broader lithological variation. Three rotated factors explain 75.35% of total variance. F1 explains the largest share (39.78%) and is most effective for characterizing lithological background, whereas F3 (17.28%) is most effective for gold targeting because Au, As, Sb, and Mo load together and its high scores form spatially coherent clusters. A robust composite index and P95 spatial grouping delineate eight targets. T1–T3 substantially reproduce conventional anomalies HS4, HS2, and HS1; T4 partly overlaps HS3; and T5–T6 are new multielement targets. Maximum Au values in HS1–HS4 are 14.9, 14.4, 49.8, and 754.3 ng/g, respectively, and each anomaly is supported by a different pathfinder suite. Removing the maximum Au value leaves the F3 loading vector nearly unchanged (r = 0.997). The strongest sedimentary Au responses are most plausibly related to upstream mineralized shear zones, weathering release of Au-bearing sulfides, and hydraulic concentration, but catchment mixing and a possible nugget effect preclude direct inference of orebody boundaries. The results provide a transparent basis for staged field verification rather than confirmation of mineralization.