Soil-order stratified pedotransfer functions for particulate and mineral-associated organic carbon in Taiwan Topsoils: Contrasting controls in Ultisols and Inceptisols

Understanding how soil organic carbon (SOC) fractions respond to environmental and physicochemical factors is critical for improving SOC modeling. However, the mechanisms underlying particulate (POC) and mineral-associated organic carbon (MAOC) accumulation, as well as the development of estimation models, remain poorly constrained in tropical and subtropical soils. This study examined 84 Ultisols and 66 Inceptisols to identify the key drivers of POC and MAOC formation across Taiwan and developed soil-order-stratified pedotransfer functions (PTFs) using stepwise multiple linear regression (SMLR). Compared to Random Forest and Cubist models, SMLR showed comparable predictive performance ( R 2 = 0.68–0.90) while revealing contrasting dominant controls, suggesting distinct sequestration mechanisms across soil orders. In Ultisols, MAOC sequestration follows a sorptive pathway constrained by mineralogy. However, mineral surfaces appear to approach a saturation threshold, limiting further C accrual. Conversely, MAOC in Inceptisols is driven by a microbial necromass pathway, where sequestration is dictated by nitrogen availability, pH, and specific surface area rather than mineral abundance. The Inceptisol models exhibited relatively lower R 2 and higher RMSE, reflecting the inherent stochasticity of younger, less-buffered profiles. Notably, the negative influence of crystalline Fe (hydr)oxides on MAOC of Inceptisols, coupled with its strong negative correlations with temperature, underscores the vulnerability of these developing soils to climate-driven C loss. Overall, our results demonstrate that soil-order stratification provides a robust and permanent structural framework for regional SOC prediction. By capturing pedogenic constraints that persist through land-use transitions, this approach establishes a stable baseline for quantifying C sequestration potential across diverse landscapes.