Soil erosion is a critical issue on the Loess Plateau due to weak soil and intense summer rainfall. Plant roots provide essential soil stabilization. A split-plot field experiment was conducted in Liulin County, Shanxi Province, to evaluate the effects of biodegradable mulch and organic amendments on maize root development and soil stabilization. The main plots included no mulch (N) and biodegradable mulch (M). The subplots comprised five treatments: control (CK, no amendment), peat (PT), biochar (BC), fermented pig manure (PM), and corn stover (CS). Correlation and principal component analyses were used to elucidate the underlying mechanisms. The results showed that organic amendments were the primary factor influencing the root and soil properties. Peat and biochar significantly raised the root surface area density (RSAD, p < 0.05) and root–soil composite cohesion (with increases of 122.56% and 109.06% for NPT and NBC compared to NCK, respectively). Biodegradable mulch, and its interaction with the organic amendments, had no statistically significant effect on either the root–soil composite cohesion or root system parameters. The strong positive correlations of cohesion with the root length density (RLD, r = 0.80) and root volume density (RVD, r = 0.81) highlight that root occupancy is the key mechanism for enhanced shear resistance. Therefore, biochar is recommended for its effectiveness in enhancing soil retention and its potential co-benefits for carbon sequestration. This study provides a technical reference for sustainable agriculture on the Loess Plateau, while also acknowledging the need for further research on long-term carbon dynamics.

Organic amendments influence several soil functions simultaneously, but whether material type matters at a fixed carbon (C) input is unresolved. In a rainfed maize field on the Loess Plateau under uniform biodegradable film, peat (MPT), biochar (MBC), fermented pig manure (MPM) and maize straw (MCS) received equal C (1200 kg C ha⁻¹) with an unamended control (MCK) in a randomised complete block design (n = 3). Seventeen indicators formed a six-dimension soil multifunctionality (SMF) index; metagenomes supplied predictors of diversity, connectivity, Y-A-S investment and C-, N- and P-cycling-gene diversity. All four raised SMF 4.0–4.9-fold without differing detectably; yet, their dimensional composition differed (MBC, carbon storage; MPM, nutrient supply and enzymes; MCS, structure; MPT, water regulation and productivity). Functions were predominantly synergistic, forming two weakly linked modules (nutrient–enzyme and carbon–water–biomass) whose separation, though weaker, persisted after treatment adjustment and the few significant negative correlations reflected the reverse orientation of bulk density rather than functional trade-offs. S/Y stress-tolerance investment and connectivity were the strongest predictors, species richness the strongest correlate of SMF (ρ = 0.66), and cycling-gene diversity tracked both SMF and nutrient supply (ρ = 0.53–0.68). Five SMF-positive genera clustered on two phylogenetic branches and the most connected module; none of the 14 hub and SMF-associated genera changed detectably in absolute abundance, yet seven shifted in relative share. At equal C input, material quality reshapes the combination, not the overall level, suggesting function-targeted material choice, with locally available straw and pig manure-favoured pending multi-year, multi-site validation.