Quantifying Soil Structure to Augment the Relevance of Laboratory-Based Soil Hydraulic Properties for Environmental Modelling

The flow and retention of water in soils affect the quality of life of a society by influencing (i) food and fiber production (via supporting plant growth), (ii) water quality (via e.g. soil erosion), and (iii) the safety of humans and their possessions (via flooding and land-slides) and impacts a multitude of ecosystem services. Mathematical modelling is perhaps the most effective tool for hypothesis testing and event-forecasting of these phenomena. Since modeling of water flow in soils is an essential component in these models, accurate values of the soil properties defining the retention and flow of water (hydraulic properties) are required. Measurement of these properties in the laboratory is time consuming and expensive; therefore, they are often estimated from basic soil properties. Current estimations of soil hydraulic properties are uncertain because: (1) measurements in the laboratory poorly represent hydraulic properties in the field; and (2) estimation approaches are often based on measurements on disturbed soil samples, while a correct representation of the hydrological processes in the field requires quantifying the undisturbed 3-D pore space of the soil. The aims of this project were to (i) quantify the complex 3-D soil pore system and, (ii) translate improved laboratory soil data into new methods for estimating soil hydraulic properties relevant at the field scale. We have collected 252 soil samples from more than 30 soil types of diverse soil- and landscapes across Norway and produced a methodologically consistent database of soil hydraulic properties that is unique in featuring e.g. soil pore-network metrics derived by X-ray imaging, the measurement of soil water retention by 4 successive methods, and currently the largest collection of continuous soil particle size data by the novel PARIO method. We have used these data to test a theoretical model that couples the complex pore-network of soil and its capacity to conduct infiltrated water. It will also be used to validate the proposed Kullback-Liebler Divergence (KLD) concept (i.e. to what extent a probability distribution differs from another) as a single-number numeric representation of soil structure effects on soil hydraulic function. Testing the KLD concept on less detailed and methodologically mixed ? therefore noisy - historic data remained inconclusive. Early exploration of correlations among the data types suggest strong predictive power of X-ray imaged soil pore metrics in inferring soil hydraulic properties. This signals that our research hypothesis will likely be confirmed and we will advance the state-of-the art in the estimation of soil hydraulic properties when the planned machine-learning models are published. We also completed two elaborate field campaigns, in which we monitored water transport ? and in one case tracer transport - in meter-scale soil volumes using a complex set of soil moisture and geophysical sensors in a 3-D configuration, followed by intensive sampling and subsequent imaging and laboratory analyses as described above. The HYDRUS 3D model, the most widely used simulation model of its kind has been set-up and is being parameterized to simulate these 3-dimensional flow fields and quantify the field-level benefits from this project. The project also helped produce and release software for a more effective handling of X-ray derived soil images, which may help improve the method?s feasibility in the future. The project provides new and improved foundations for modelling environmental processes by its core products and findings, while it also enhances future research potential via the engagement of students and young scientists and by inspiring new research directions.

We built two 3-D field data collections of heterogeneous soil water transport patterns and produced a unique, methodologically consistent database of soil hydro-physical data, complete with X-ray imaged soil pore-network metrics. The project will benefit science by proposing - to our knowledge still the first - machine-learning based empirical estimation model of soil hydraulic properties based on X-ray imaging of soils, and by that, improving the estimation of field-effective soil hydraulic properties. We are in process of filling a knowledge gap in bio-physical modeling by validating a single-metric that describes the effect of soil structure on soil water retention properties. We contributed to the (semi-)automatization of X-ray image analysis, making the use of the technique itself more feasible. The project will impact society at large indirectly through the improvement of model-data support for future scenario studies, risk assessment and environmental planning.

Successful simulation-based environmental studies and relevant mapping applications rely on accurate predictions of soil hydraulic parameters from readily available soil properties. Progress in the latter area has been stagnant for at least 10 years - the base data used in such estimations has not changed much in over 30 years - which have compromised our ability to simulate environmentally relevant processes. This project has identified two reasons for this: (1) soil hydraulic properties measured in the laboratory poorly represent hydraulic properties at the field scale for reasons that are not yet well understood; and (2) conventional approaches to estimate soil hydraulic properties are primarily based on properties of the solid constituents of disturbed soil samples, while hydrological processes in the field are governed dominantly by the 3D void system of the undisturbed, structured soil. The aims of this project are to advance the scientific frontier on each of these identified areas by contributing to i) the understanding and quantification of 3D soil pore geometry; and ii) translating laboratory soil data into field-effective soil hydraulic properties. Existing international data as well as extensive dual-scale data collection campaigns will serve as the necessary soil data and validation on field moisture regime. Novel mathematical concepts, non-invasive geophysical measurements, and X-ray computed tomography imaging will be coupled with traditional measurements to generate the necessary data pool. Cutting edge machine learning tools will be used to quantify data relationships and the predictive power of the new types of data, and 2D/3D simulation modelling will be used to quantify the benefit from the new findings in the context of 2 selected field-sites, using existing and ongoing field data collection campaigns. The study promises to provide new, improved foundations for parameterizing environmental studies.