Summary
This study demonstrates the application of Random Forest ensemble learning to estimate root zone soil moisture at daily timescales across an agricultural catchment, comparing data-driven machine learning performance against conventional process-based hydrological models. Random Forest showed slightly higher accuracy for interpolation and comparable accuracy for extrapolation, with a key advantage being independence from soil hydraulic parameters—making it particularly valuable for data-poor regions where such parameters are incomplete or unavailable. However, poor predictive performance under extreme moisture conditions highlighted limitations in learning from infrequent events and incomplete representation of subsurface processes in the model's input features.
UK applicability
The methodology could be applicable to UK agricultural regions where detailed soil hydraulic parameters are lacking or expensive to obtain, particularly for operational soil moisture monitoring. However, UK's variable and extreme precipitation patterns would require careful validation to ensure adequate representation of the full range of wet and dry conditions in training datasets.
Key measures
Root zone soil moisture (RZSM) prediction accuracy; interpolation and extrapolation performance; model comparison with process-based approaches; performance under extreme wet and dry conditions
Outcomes reported
The study evaluated Random Forest machine learning algorithm performance for estimating root zone soil moisture (RZSM) on a daily timescale over a 2-year period (2016–2018) at an agricultural catchment using in situ measurements. RF predictions were compared against process-based model simulations combined with data assimilation for both interpolation and extrapolation accuracy.
Topic tags
Dig deeper with Pulse AI.
Pulse AI has read the whole catalogue. Ask about this record, its theme, or how the findings apply to UK farming and policy — every answer cites the underlying studies.