Evaluating the performance of percolation-based effective medium approximation and sigmoidal model for modeling thermal conductivity of tropical soils

Soil thermal conductivity (λ) is essential for understanding coupled matter and energy flux in porous systems. However, its prediction has remained unsettled due to its intricate relationship with soil properties such as soil moisture content (θ) and texture. This study investigates λ–θ relationship in 15 tropical soils from Ghana. The results show that soil λ–θ relationship is strongly nonlinear and depends on texture with fine-textured soil exhibiting a flat tail in λ the dry range. Four distinct domains, namely, hydration, pendular, funicular, and capillary, ranging from dry to wet condition, were identified. Two modeling approaches, percolation-based effective medium approximation (P-EMA) and an empirical sigmoidal model (SM), were evaluated on the λ(θ) dataset. An explicit λ(θ) P-EMA variant outperformed a θ(λ) variant, achieving performance on par with the SM. However, the explicit λ(θ) P-EMA performs well in the hydration regime, whereas the SM performs well in the capillary regime. The explicit λ(θ) P-EMA and SM had difficulty fitting the nonlinear behaviour at intermediate θ beyond the pendular domain, possibly due to the structural complexity, which is also linked to phase distribution. This study advances our knowledge on λ–θ relationships in Ghanaian soils.