Uncertainty in soil hydraulic properties limits the predictability of global water and carbon dynamics

Water transport at the land surface and in the soil – the critical zone - is highly dependent on the soil hydraulic properties. Such properties influence simultaneously the terrestrial water and carbon cycles as they determine the water fluxes in the soil and the soil’s water holding capacity, ultimately affecting runoff production, groundwater recharge, and the amount and temporal variability of plant available water (i.e. plant water stress). Despite their paramount importance, limited global information concerning the spatial distribution of soil hydraulic properties currently exists. Information at the global scale, commonly used in Earth System Models, mostly originates from pedotransfer functions (PTFs). PFTs are empirical relations that express the dependence of soil hydraulic properties on easily measured attributes as soil properties. Several PFTs currently exist, which adopt different formulations, spanning from simple linear regressions to elaborate machine learning, and are trained with different datasets, yielding different soil hydraulic properties for the same soil texture.   

The question we ask in this study is: how does uncertainty in the soil hydraulic parameters propagate in global ecosystem responses? To achieve this, we deploy a numerical experiment covering many different ecosystems. The terrestrial ecosystem model T&C is used to model energy, water, and carbon dynamics at 80 locations worldwide, spanning all climatological regimes, major biomes and soil types. Soil hydraulic properties at each site were estimated using six widely used PTFs starting from local soil textural information. Uncertainty propagation from soil hydraulic properties to modelled ecosystem dynamics was evaluated for all sites and its dependence on soil textural properties and local topography was quantified.

Our results highlight that uncertainty propagation from hydraulic properties to ecosystem dynamics is much stronger for hydrological fluxes (e.g. infiltration, groundwater recharge and runoff production) than carbon dynamics (e.g. gross and net primary productivity and leaf area dynamics) or energy fluxes (net radiation, sensible and latent heat). Uncertainty in hydrological fluxes can be up to 400% using different PTFs, whereas uncertainties in carbon and energy fluxes are typically less than 20%. The largest uncertainties were observed for slow draining soils, containing large fractions of clay, located in regions with intermediate values of wetness (i.e. annual precipitation ≈ annual potential evapotranspiration). Complex topographic features further enhance the role of uncertainty in soil hydraulic properties. Lateral water redistribution affects both runoff production and soil moisture dynamics increasing the effects on both hydrological and carbon dynamics.