Vertical Distribution of Forest Soil Physicochemical Properties and Their Environmental Covariates in Maoming City, Southern Subtropical China

Forest soil organic carbon (SOC) and nutrient dynamics are critical for terrestrial carbon cycling, yet full-profile soil property characterization at regional scale remains limited in southern subtropical China. Based on 0–100 cm soil profile data from 443 forest sampling points in Maoming City—6645 field observations aggregated to 2215 point × layer records for statistical modelling—this study investigated vertical distribution patterns of SOC, total N, total P, total K, pH, and molar stoichiometric ratios, and quantified the relative explanatory contributions of stand type and topographic factors using linear mixed-effects models (LMMs), random forest (RF), and geostatistical kriging. SOC and total N exhibited pronounced surface accumulation, declining by approximately 50% and 44% from the surface (12.75 g kg−1 and 941 mg kg−1) to the 80–100 cm layer (6.35 g kg−1 and 530 mg kg−1), respectively. Total P showed minimal vertical variation (≈5%), total K remained broadly stable, and pH increased slightly with depth (4.55 → 4.66). Stand type was significantly associated with SOC, total N, total K, pH, and most stoichiometric ratios (all p < 0.05); surface SOC was higher in broadleaf mixed and coniferous forests than in eucalypt, economic tree species, and litchi/longan plantations. All stoichiometric ratios declined significantly with depth. RF models showed that soil depth had the highest permutation importance for SOC and total N (0.44–0.59), while stand type ranked first for pH and total K (0.40–0.49); however, out-of-bag R2 values were modest (–0.06 to 0.23), indicating that substantial variation remains unexplained by the four measured predictors. All properties exhibited moderate spatial autocorrelation (nugget ratio 0.48–0.67). Total K showed moderate interpolation accuracy under random leave-one-out validation (R2CV = 0.458), whereas spatially structured block cross-validation yielded negative R2 for four of the five properties; the kriging maps accordingly visualise the spatial arrangement of the measured properties. These results provide regional baseline evidence for 0–100 cm nutrient vertical patterns and soil carbon density in southern subtropical China.