Smarter sediment management: forecasting tools for sustainable dredging
Type:
Presented during:
CEDA Dredging Days 2026
Authors:
A. Minnigin, P. Guard, C. Earlie, S. Maroju, K. Ghaly, T. Matthweson, N. Morarji and A. Aldrich
Abstract
Maintenance dredging at busy ports is costly and carbon-intensive, yet current practice relies on constrained dredging schedules with limited forecasting. We present a novel predictive dredging that uses advances in data processing, machine learning, and process based modelling to support proactive and adaptive sediment management decisions. The novel approach is demonstrated through two pilot studies: one using a data driven statistical approach and one using a machine learning emulator trained on outputs from a calibrated physics-based model. In the Houston Shipping Channel, publicly available hydrographic and metocean datasets were used to train timeseries and neural network models across 189 transects. Both models reproduced group mean sedimentation trends within roughly one metre of observed values, with best performance in transects that showed stable sedimentation behaviour. Comparative evaluation showed forecasting errors on the order of 36-38% sMAPE for the most predictable regions. The results also highlight the benefit of using both model types together. Neural Networks capture nonlinear and event driven responses, while ARIMA models provide transparent links to physical drivers that support user confidence. At the Port of Townsville, a fully coupled three dimensional hydrodynamic, wave, and sediment model generated a physics consistent annual sedimentation dataset. This supported scenario testing of alternative harbour layouts and provided rich training data for machine learning emulators. Initial Random Forest emulation achieved R? values between 0.60 and 0.78 across dredge management zones, demonstrating potential for rapid prediction without the need to run computationally expensive solvers. Together, these results indicate how predictive dredging can shift port maintenance from reactive to anticipatory practice, reduce unnecessary dredging, and support more sustainable and timely waterway management.