Soil Structure and Land Surface Controls on Soil Hydraulic Properties and Processes: Applications of Machine Learning, Unmanned Aircraft Systems, and Observations from Long-Term Conservation Agriculture Management
Abstract
Soil moisture exerts a strong influence on ecology and energy balance of the environment. Its high variability in space and time, however, makes it difficult to measure accurately. Predicting the dynamics of soil moisture is also challenging because of the complex relationship between soil hydraulic properties and other physical characteristics. In this study, the links between soil structure and land surface characteristics with soil hydrology were determined using machine learning, unmanned aircraft system (drone), and observations from long-term conservation agriculture management.
The saturated hydraulic conductivity is an essential property of soil that determines the fate of soil moisture. Machine learning models were developed that predict saturated hydraulic conductivity with significantly improved accuracy than current models. The impact of soil structural changes on hydraulic conductivity was further investigated using these models. Soil moisture across a grassland surface soil was interpreted from multispectral imagery taken by a drone using machine learning-based methods. The interpretations had reasonable accuracy and were used to describe how land surface variables relate to surface soil moisture across a landscape. Soils that have been under different tillage and cover crop systems for the past 18-years were studied and compared to determine the implications of management on soil structure. Conventional measures of soil hydraulic properties and numerical simulations were applied to investigate the implications on soils' capacity to capture and store water.