Going deeper with Rijkswaterstaat
Even in complex operations, organizations rely on manual skills as well as tech. And with teams often working at full capacity, we set out to explore how machine learning could boost insights without increasing the volume of data gathering.
Rijkswaterstaat provides daily information on where the shallowest depths are in Dutch rivers – so that boats and vessels can navigate safely. Inspectors measure depths in rivers by heading out to the locations where experience tells them the shallowest parts will lie.
The work was done with Rijkswaterstaat Datalab and analyzed river depth data, water levels and discharges measured over a 100km stretch of the river Waal.
The model approach
Models were built that can predict how the bottoms of rivers evolve through the day.
Human experience and instinct are reliable but what our data science provided was the ability to process a lot of complex data quickly and build predictive, objective real-time models that describe the behavior of the river depth across the 100km stretch, up to several days into the future.
Our models can also point the measurement fleet to places that are measured less frequently and even find an even shallower spot that might go unnoticed, helping to maintain the navigation safety.
These systems can continue to learn from changing rivers and can alert Rijkswaterstaat if anomalies are detected.
The conclusion
The models built are continuously improving their performance with each measurement provided to them, and they do not require any additional effort on the inspectors’ part.
The work again demonstrates how machine learning can augment human-based know-how to improve efficiency and free up resources.