Detection of attenuations in watercourses

Wetterskip Fryslân

The algorithm sees from aerial photos in which watercourses there have been attenuations.

Last change on 12th of March 2025, at 8:13 (CET) | Publication Standard 1.0

• Publication category: Other algorithms
• Impact assessment: Field not filled in.
• Status: In use

General information

Theme

Nature and Environment

Begin date

2023-11

Contact information

[e-mail]

Responsible use

Goal and impact

By using the algorithm, we can detect more/faster illegal attenuations. Good water management in the Netherlands is necessary to protect our country from flooding and an excess or shortage of rainwater, for example. A good water level is important for nature conservation, recreation, fishing and water transport, among other things. It is important to detect illegal attenuations as they can disrupt the water system.

Considerations

By deploying this algorithm, we as a water board can detect illegal attenuations faster, allowing us to take quicker action to restore the original situation. This contributes to a better functioning water system.

By using the algorithm, we work more efficiently because we can detect more attenuations in a short time. Where previously a Wetterskip employee searched for attenuations using an aerial photograph, now the computer does this. The algorithm works on the basis of colour recognition and can thus identify attenuations. As a result, the algorithm only works as a detection tool, as an employee manually checks the computer's results. Besides the visual check on a monitor, a field employee visits the location of the possible attenuation. Therefore, there is always human supervision of the algorithm's operation.

Human intervention

An employee of the water board checks if the observed attenuations are correct.

They then check whether permits have been issued for any found attenuations, because if so, the attenuation is not illegal.

Then an employee in the field checks the attenuations found to be illegal. Where necessary, immediate action is taken by enforcement.

Through this human intervention, we know whether the algorithm continues to work properly and, if necessary, train the algorithm when it shows anomalies.

Risk management

With human supervision. There is a working description. Measures are included in the DPIA.

Legal basis

Water board regulation

Links to legal bases

Waterschapsverordening: https://lokaleregelgeving.overheid.nl/CVDR705929/

Operations

Data

Aerial photos and Basic Large Scale Topography (BGT).

Links to data sources

PDOK: https://app.pdok.nl/viewer/

Technical design

The model learns to recognise water on aerial photographs. It looks at the colours on the photos and uses them to determine if there is water. For example: A very dark colour, almost black, may indicate the presence of water. By practising with photos that show water, the model learns what water looks like.

When the model has finished learning, we can use it to see where water is on aerial photographs. That way, we can see exactly where ditches are. Then we compare what the model has found with existing data about water. This allows us to see if anything has changed. We check the changes we have found manually to make sure they are attenuations.