Bimodal PDI modeling improves parameter estimation and uncertainty reduction of hydric soils on the Tibetan Plateau Authors

Hydric soils on the Tibetan Plateau play a crucial role in regulating alpine ecohydrological processes, yet their hydraulic properties remain insufficiently characterized due to strong heterogeneity and dual-domain pore structures. In this study, a total of 45 undisturbed and 45 disturbed soil samples were collected from nine representative sites, and evaporation and multi-step outflow experiments were performed to obtain soil-water retention curves (SWRCs) and unsaturated hydraulic conductivity functions. The experimental data were fitted using unimodal van Genuchten (VG), unimodal Kosugi, and bimodal particle discrete interaction (PDI) models, with parameter estimation conducted through Bayesian inference. Results demonstrate that the bimodal-PDI scheme significantly improved fitting performance for sphagnum-rich and structurally complex soils, increasing R² from 0.91-0.92 to 0.96, reducing RMSE by 30-40%, and achieving near-identical corrected Akaike information criterion (AICc) values compared to unimodal models. More importantly, Bayesian posterior analysis revealed that the bimodal-PDI reduced parameter uncertainty by approximately 40% and minimized parameter correlation, thereby enhancing robustness in hydraulic parameterization. Mechanistically, the bimodal-PDI approach effectively captured macropore-driven near-saturation drainage and micropore-controlled residual retention. These advances provide a quantitative basis for improving soil hydraulic representations in land surface and climate models, offering new insights into water balance simulation and ecological management in the "Third Pole" region.