Soil structure is an important omission in Earth System Models

The study addresses the omission of soil structure in the representation of soil hydraulic properties within Earth System Models. Soil is central to land–atmosphere–biosphere coupling, controlling energy and water exchanges, vegetation productivity, and extremes. Current large-scale models typically estimate hydraulic properties via pedotransfer functions largely based on soil texture, neglecting biotically induced soil structural features such as aggregates, biopores, and macropores. This is partly due to biases in soil sampling (focus on agricultural soils and avoidance of roots/macropores) and challenges in quantifying structure effects even at profile scale. The authors hypothesize that failing to account for soil structure biases hydraulic parameterization and may significantly affect infiltration-runoff partitioning, groundwater recharge, energy fluxes, and potentially climate. They propose a simple, first-order parameterization to include biophysical soil structure effects and assess impacts at ecosystem and global scales.

Pedotransfer functions (PTFs) trained on limited, geographically clustered datasets (often agricultural soils) underpin most ESM soil parameterizations. These PTFs, based primarily on texture, underrepresent structural effects linked to biological activity, leading to biased hydraulic properties for natural ecosystems (forests, undisturbed soils). The literature documents the influence of soil structure on hydraulic functions and ecology, yet few PTFs incorporate structure. Databases like UNSODA show saturated hydraulic conductivity distributions that likely miss high-conductivity tails associated with structural macroporosity. Attempts to include soil organic matter in PTFs only partially reduce bias; simple predictors (bulk density, soil organic matter) have low predictive skill for Ksat. Observations indicate Ksat can be one to three orders of magnitude larger when structural effects are present compared to texture-only estimates, with positive correlations between Ksat and macroporosity or soil organic carbon, and between vegetation productivity and infiltration/hydraulic conductivity. Abiotic contributors to structure (e.g., shrink-swell, freeze-thaw) are recognized but hard to generalize at large scales and often seasonal.

The study introduces a first-order parameterization linking biotic soil structure to vegetation productivity and integrates it into two modeling frameworks: the ecosystem model Tethys-Chloris (T&C) for local-scale analyses and the global climate model OLAM for global-scale impacts.

• Soil hydraulic functions: Base parameterization uses van Genuchten–Mualem (VG) functions for water retention and hydraulic conductivity. A composite approach conceptually represents structural and textural domains. For efficiency in land-surface models, most simulations modify only the hydraulic conductivity function near saturation to account for structure while keeping the retention curve unchanged (θ_mac = 0), acknowledging that structural macroporosity primarily increases near-saturated conductivity.
• Structure parameterization: Biotic structure intensity is proxied by local Gross Primary Production (GPP). The ratio K_sat/K_sat,tex scales linearly from 1 at GPP=0 to 1000 at GPP=3000 gC m⁻² yr⁻¹, consistent with literature extremes. The depth dependence scales with cumulative fine root biomass, maximizing at the surface and reducing to 1 below rooting depth. Additional parameters (e.g., α_sat/α_tex) are linked to K_sat/K_sat,tex using sparse literature fits and set within plausible bounds; sensitivity to α is low. In OLAM, structure effects are applied as a correction to Ksat that diminishes as water potential drops from 0 to −10 cm to reflect macropore drainage.
• Local ecosystem-scale experiments (T&C): Twenty sites across diverse climates/biomes were simulated with hourly observed meteorology (3–31 years per site). Three soil scenarios: (i) ORI: locally tuned Saxton–Rawls PTF-based parameters; (ii) VG: global SoilGrids-250m textures converted via Tóth PTFs, no structure; (iii) VG+SS: same as VG but with soil structure parameterization applied to conductivity, scaled with root depth. Outputs included long-term averages of runoff, deep drainage (recharge), ET components, energy fluxes, GPP, LAI, and vertical soil moisture profiles.
• Global simulations (OLAM): Two 35-year simulations with 200 km atmosphere and 50 km land-surface grids (except Antarctica coarser) were performed after a century-long soil moisture/groundwater spin-up. Scenarios: without structure (NSS) and with structure (WSS). The first 5 years were discarded; 30-year means were analyzed. Eleven climatic/land-surface variables were tested across 27 regions and 12 months using two-sample t-tests to assess statistical significance of differences (WSS−NSS), accounting for internal climate variability. Maps of differences for key variables were produced.

• Local/ecosystem scale: Including soil structure (VG+SS vs. VG) substantially alters hydrologic partitioning in productive and wet ecosystems and in fine-textured, poorly drained soils. • Example: Tropical rainforest (Manaus) experienced +1050 mm/yr deep drainage (recharge) and −1280 mm/yr surface runoff after including structure, about 40–45% of annual precipitation. • Across 20 sites, median changes were approximately +46 mm/yr drainage and −48 mm/yr runoff, nearly balancing at the median. • Even small absolute changes can be large relative to recharge in semi-arid systems (e.g., 3 mm/yr equals ~17% change at Short Grass Steppe). • Energy fluxes, transpiration, GPP, and LAI changes were generally small (<2% vs. VG), with two sites showing latent heat changes up to ~15%. Differences are typically larger (5–10%) when comparing VG+SS to locally tuned ORI parameters, underscoring uncertainty from using global-texture-derived parameters. • Soil moisture vertical profiles often showed higher water contents at depth under VG+SS due to increased vertical redistribution during wet periods; near-surface moisture was largely unaffected due to evaporation and root uptake. Semi-arid or sandy sites showed minimal profile changes.
• Global scale: Differences between WSS and NSS were generally small and statistically insignificant. • Near-surface air temperature differences were mostly within ±0.4 K; latent heat, vapor pressure, and precipitation showed minimal systematic differences. • Out of 3564 tests (11 variables × 27 regions × 12 months), 158 (4.4%) were significant at α=0.05 and 43 (1.2%) at α=0.01, consistent with expected Type-I error rates, indicating no robust global-scale signal. • A few localized patterns suggested reduced latent heat and slightly higher temperatures under WSS, but statistical significance was weak.
• Interpretation: The lack of a global signal is attributed to coarse model resolution and smoothing of rainfall intensity, which fail to trigger near-saturation processes and lateral redistribution where structure matters most.

The study’s results confirm the hypothesis that soil structure significantly influences hydrologic partitioning at the ecosystem scale, particularly enhancing infiltration and deep percolation while reducing surface runoff in productive, fine-textured soils. However, anticipated cascading impacts on energy fluxes and climate at the global scale were not detected in current ESM configurations. This discrepancy is likely due to model resolution limits that suppress key activation mechanisms for soil structure effects: intense, short-duration rainfall and fine-scale lateral redistribution of water across heterogeneous topography. Consequently, despite meaningful local impacts on recharge and runoff, the aggregated global climate response remains elusive. The findings also highlight that uncertainties from PTF choice and reliance on global soil maps can be comparable to or greater than the effect of adding soil structure for some fluxes, suggesting the need for improved soil parameter data and methods. Incorporating structure offers a principled alternative to ad hoc tuning practices and may be critical for regions and processes where runoff–recharge partitioning affects water resources and feedbacks to the atmosphere, especially when models resolve lateral flows and extremes.

The paper introduces a simple, systematic parameterization that links biotically driven soil structure to vegetation productivity (GPP) to modify near-saturated hydraulic conductivity in land-surface and climate models. Implemented in T&C and OLAM, the approach shows substantial local effects on infiltration–runoff partitioning and deep drainage, with modest impacts on energy and carbon-related fluxes. At current global modeling resolutions, soil structure effects do not produce a statistically robust climate signal. The work underscores a substantial omission in standard PTF-based parameterizations and provides a framework for integrating structure effects while calling for higher-resolution modeling that captures intense rainfall and lateral hydrologic processes. Future research should: (i) conduct catchment- to regional-scale studies with explicit lateral flows to assess spatial propagation of structure effects; (ii) improve parameterization using soil class, parent material, qualitative structure descriptors, and new quantitative datasets; (iii) incorporate abiotic structural drivers and mechanistic models of bioturbation and aggregation; and (iv) investigate implications for soil biogeochemistry and the carbon cycle via altered soil moisture profiles.

• Coarse spatial resolution in global simulations (∼50 km land, ∼200 km atmosphere) limits representation of lateral hydrology, stream networks, topographic controls, groundwater upwelling, and extreme rainfall intensities, suppressing activation of structure-driven processes.
• Structure parameterization is first-order and uncertain: linear scaling of K_sat/K_sat,tex with GPP up to 1000, limited empirical constraints, and no dependence on soil texture or land use despite evidence of differential effects.
• Abiotic structural processes (shrink-swell, freeze-thaw) are not included; seasonal dynamics are neglected.
• In most simulations, structure impacts on water retention were ignored (θ_mac=0) for computational efficiency; dual-porosity/non-equilibrium effects were not represented.
• OLAM implementation reduces structure effect rapidly as matric potential declines (0 to −10 cm), a simplification of macropore drainage behavior.
• Reliance on global soil maps and PTFs introduces uncertainties; local tuning in ORI highlights the variability and potential bias in texture-derived parameters.
• Limited observational datasets explicitly separating textural and structural hydraulic parameters constrain calibration/validation of the parameterization.