A regime-based piecewise function for predicting gas diffusion coefficients in repacked soils

• A piecewise D s / D 0 model is developed based on three distinct soil water regimes. • The model captures dynamic pore connectivity ignored by traditional methods. • Key parameters are estimated from soil texture and porosity. • Validation shows a 21.1–61.6% reduction in RMSE compared to existing models. • The model enables more accurate predictions of greenhouse gas fluxes. Soil gas diffusion coefficient ( D s ) is a critical parameter governing gas transport in unsaturated soils, affecting climate regulation, contaminant fate, and ecosystem functioning. Traditional models estimate the relative gas diffusion coefficient ( D s / D 0 , dimensionless, where D 0 is the gas diffusion coefficient in free air) as a function of air-filled porosity, neglecting changes in pore connectivity across moisture regimes. This study presents a novel regime-based model that partitions D s / D 0 into three domains derived from soil water retention characteristics: (1) a near-saturation regime, where gas diffusion is strongly inhibited due to disconnected air-filled pores; (2) a capillary-dominated regime where D s / D 0 decreases exponentially with decreasing air-filled saturation ( S a ); and (3) an adsorption-dominated regime where D s / D 0 decreases linearly with S a . The model integrates soil texture, porosity, and critical water thresholds, calibrated using D s / D 0 data from 48 soils (Soils 1–30, Table 1) spanning a wide range of textures and porosities. Validation with independent datasets (Soils 31-51c, Table 2) yielded a mean root-mean-square error (RMSE) of 0.019 for D s / D 0 estimates, representing a 21.1–61.6% improvement over established models. This improvement is particularly notable in capturing regime-dependent diffusion behavior. By advancing the predictive capability for gas transport in variably saturated soils, this framework facilitates more accurate modeling of greenhouse gas fluxes and contaminant volatilization.