Comment on “On the equivalence between laboratory and field estimates of soil water retention parameters and its implications for modeling oxygen diffusion in structured agricultural soils” by H. Yadav et al.

Yadav et al. (2025) compared soil water retention parameters (SWRPs) derived from laboratory and in situ measurements for structured agricultural Ultisols and examined implications for modeling oxygen diffusion. Their erratum (Yadav et al., 2026) corrected the in situ soil moisture retention curves and some fitted parameters. Using the corrected data, we point out a methodological issue in parameter estimation for field-based water retention curves: when the saturated water content θ s θ ₛ is optimized without constraints, the fitted value can become smaller than the maximum observed volumetric water content, that is, max ( θ obs ) max (θobs) in the same dataset. This seems physically inconsistent, at least for their data. We show that the inconsistency can be avoided by imposing simple constraints while maintaining or improving the fit quality. We further illustrate that triple-porosity functions can better represent the wet-end shape typical of structured soils. In the corrected Appendix A3 of Yadav et al. (2026), several dual-porosity (DP) fits to field-based curves yield optimized θ s θ ₛ values that are clearly smaller than max ( θ o b s ) max( θ obs ) . This tends to occur when θ θ decreases steeply with a small increase in suction near saturation and only a few near-saturated points are available. If θ s θ ₛ is left unconstrained, a nonlinear optimizer can then reduce RMSE by lowering θ s θ ₛ below the observed θ θ range, thereby improving the numerical fit at the cost of physical consistency. To assess whether this behavior reflects a limitation of the retention model itself or of the optimization setup, we requested and obtained from the authors the in situ h- θ θ data underlying the corrected Figure 4b and Appendix A3. Using these data, we re-fitted the field-based soil moisture retention curves for all 10 plots reported in the corrected Appendix A3 with the open-source Python program unsatfit of Seki et al. (2023). We did set θ r = 0 θ ᵣ = 0 and fitted several multimodal retention functions from Seki et al. (2022): a dual-van Genuchten model (dual-VG), which is identical to the DP formulation used by Yadav et al. (2025), and two triple-porosity models, tri-VG and BC1VG2VG3 (further referred to as BVV). The latter model combines a Brooks–Corey macropore component with two van Genuchten components for the finer pores. Below, we present samples 301 and 404 as representative examples to illustrate the effects of constraining θ s θ ₛ and introducing a third pore domain. Because the choice of θ s θ ₛ is central to the wet-end mismatch emphasized in the original paper, we examined two physically motivated constraints for field-curve analyses. In one set of refits, we fixed θ s θ ₛ to max ( θ o b s ) max( θ obs ) for each in situ dataset. In a second set of refits, we fixed θ s θ ₛ to the laboratory-fitted θ s θ ₛ value reported for the corresponding sample by Yadav et al. (2025), which provides a laboratory-consistent estimate of the saturated water content for the same material. For sample 404, when using the laboratory-fitted θ s θ ₛ , the retention data do not cover the interval between θ s θ ₛ and the wettest in situ observation. This causes the macropore subfunction to primarily bridge a data gap, with its parameters becoming weakly identifiable. In that case, consistent with the approach discussed for near-saturated gaps in trimodal fitting (Seki et al., 2026), we fixed the wet-end macropore-scale parameter as h b 1 = 1 / α 1 hb1 = 1/α ₁ = 0.01 m for tri-VG and BVV to obtain stable, physically reasonable curves. For sample 301, the near-saturated decline is better supported by data, and no such fixing was required. Table 1 summarizes the saturated water contents of two representative samples. For both samples, the in situ fitted θ s θ ₛ reported by Yadav et al. (2026) is smaller than max ( θ o b s ) max( θ obs ) , whereas the laboratory-fitted θ s θ ₛ is substantially larger than max ( θ o b s ) max( θ obs ) . This comparison shows that the unconstrained in situ optimization did not simply converge toward a higher θ s θ ₛ closer to an independent saturation estimate but instead moved in the opposite direction, below the data range. When θ s θ ₛ is constrained to the observed range by fixing θ s = max ( θ o b s ) θ ₛ = max( θ obs ) , the dual-VG formulation reproduces the corrected in situ curves with comparable or better RMSE values than the original DP fits (Table 2). This indicates that the low θ s θ ₛ values in Appendix A3 arise from allowing θ s θ ₛ to drop below the data range during unconstrained optimization, not from a failure of the DP formulation itself. In addition, introducing a third pore domain using tri-VG or BVV often dramatically improves the fit quality and better captures the characteristic shape of structured soils, namely a sharp θ θ reduction at low suctions followed by a more gradual decline at higher suctions. In the laboratory-consistent θ s θ ₛ refits, the tri-VG and BVV remain competitive. The additional flexibility is particularly useful for representing the wet-end behavior while honoring a realistic saturated water content. Figure 1 shows refitted multimodal retention curves for samples 301 and 404. The upper panels use θ s = max ( θ o b s ) θ ₛ = max( θ obs ) , and the lower panels use the laboratory-fitted θ s θ ₛ . The triple-porosity models provide single continuous curves that honor the fixed θ s θ ₛ values and measured suctions. In these examples, they capture the steep wet-end behavior more naturally than the dual-VG model. This is consistent with structured agricultural Ultisols having distinct macro-, meso-, and matrix pore domains, with trimodal formulations being effective for describing such domains (Seki et al., 2026). The differences noted above matter for applications. Yadav et al. (2025) emphasized that SWRPs affect oxygen diffusivity estimates via the Moldrup et al. (2000) model. Because this type of formulation depends on air-filled porosity derived from θ s θ ₛ and θ θ data, underestimating θ s θ ₛ in the wet range reduces the air-filled porosity and thus tends to underestimate gas diffusivity near saturation. Our refits demonstrate that it is possible to represent the full h- θ θ relation, including the wet end, without accepting θ s θ ₛ values below the observed θ θ range, thus maintaining or improving the fit quality. In summary, the corrected in situ dataset of Yadav et al. (2026) clearly supports the use of multi-porosity retention functions for structured agricultural Ultisols and other porous media. But the analysis also shows that unconstrained estimation of θ s θ ₛ can yield physically inconsistent field-based fits when near-saturated data are sparse. Fixing θ s θ ₛ to at least max ( θ o b s ) max( θ obs ) , or alternatively to an independently supported saturation estimate such as the laboratory-fitted θ s θ ₛ value for the same or a similar sample, removes this inconsistency without sacrificing fit quality. When θ ( h ) θ ( h ) exhibits a sharp wet-end decline followed by a more gradual drainage trend, triple-porosity retention functions such as the tri-VG and BVV models provide a practical way to represent distinct macro-, meso-, and matrix pore domains within a single continuous θ ( h ) θ ( h ) relation, thus improving the optimization quality relative to dual-domain formulations. K. Seki: Conceptualization; data curation; formal analysis; investigation; methodology; software; validation; visualization; writing—original draft. M. Th. van Genuchten: Conceptualization; methodology; project administration; validation; writing—review and editing.