About the Tory Channel Current and Tide Model
Background
The strong currents and turbulent flows that characterise the entrance to the Tory Channel / Kura Te Au (New Zealand) are a well-known challenge to shipping.
The flows are dominated by tides, which can be predicted with high accuracy but there is often a non-tidal component present. This non-tidal effect leads to a high level of uncertainty about the flow conditions at any given time and increases the difficulty for safe transit planning. To assist maritime navigation there is a need to supply timely, accurate and up-to-date current and tide telemetry information.
To address this requirement, over the last three years, Marlborough District Council (MDC), in collaboration with OceanNum Ltd and iXBlue, embarked on a groundbreaking project to create a real time prediction of tidal and non-tidal flows to improve the navigational safety at the entrance to the Tory Channel/Kura Te Au, New Zealand.
Measurements and the machine learned model
Current flows within the Tory Channel are driven by water level gradients. By blending high-resolution hydrodynamical model simulations and real-time water level observations, the prediction of currents at any location within the Channel has been implemented.
The system includes two tide gauges strategically placed at opposite ends of the Tory Channel, providing data at 10-minute intervals, and allowing water level gradients to be measured in near real time. From this data, the OceanNum machine-learned algorithm has been trained from the hydrodynamical model, and then used to accurately infer the flow regime from real-time measurements.
The raw water level data require processing to isolate the tidal and non-tidal signals, with the latter undergoing further reduction of noise from a signal to produce stable gradients within the Sounds.
The operational system has three core components.
- Real time observations - accessed from the MDC network of tide and weather stations.
- Data analytics engine - cloud-based preprocessing of the measured data, application of the ML algorithms, and archiving of the outcomes.
- Products pipeline - postprocessing of resultant data and production of graphics and text data for dissemination via the MDC information portals
System validation and display
To assist with mathematical model validation, Cawthron installed an Acoustic Doppler Current Profiler (ADCP) near to the Tory channel entrance. Over a 60 day period (2x30 days), the ADCP collected tidal flow data through a 30-meter water column. Comparing both sets of results, the AI model tracked well within tolerance of the measured ADCP values.
The resultant modelled tidal flow has been mapped to an overlay of the Marlborough Sounds region. The direction and numerical strength of the predicted current is represented by a graphical arrow for the tidal point. The display defaults to showing real-time current values and provides the ability to roll forward to the predicted AI values.
The modelled tidal flow has been overlaid onto a map of the Marlborough Sounds region. The tidal flow is represented by a graphical arrow which illustrates the direction and numerical strength of the predicted current. The default display shows real-time current values and allows users to transition forward to the predicted, AI values generated.
