Adopted in 2000, the Water Framework Directive requires all EU member states to protect and improve the water quality in rivers, lakes, groundwater, transitional and coastal waters, and to protect against pollution and deterioration. But in order to protect and improve the water quality, we first need to assess and understand the current conditions of our rivers and waterways – which is a difficult task. Traditionally this mapping exercise would be done ‘manually’ – by someone literally walking alongside the body of water and making a visual assessment. It was clear this work needed development and the implementation of the Water Framework Directive has seen the emergence over 73 different methodologies used across EU member states. Although promising in terms of much improved ways to manage and protect our waterways, this proliferation of methods indicates a need for a more standardised approach.
Image pattern recognition techniques – taking images via Unmanned Aerial Vehicles (UAVs) – have been successfully used to map the waterways and identify certain physical features. But the accuracy and reliability of this method depends on both the quality of the aerial images and the pattern recognition technique used. Recent studies have proved the potential of UAVs to increase the quality of the imagery by capturing high resolution photography. Combine those high resolution images with Artificial Neural Networks (ANN), which are computers, designed to work like the human brain, and you have a high precision tool that can be used to automatically recognise environmental patterns.
This kind of method enables us to obtain aerial images to a resolution as small as 2.5cm, and we recently tested this along a stretch of the River Dee in Wales. This is a significant step forward as previous imagery was at a resolution of generally 12.5 cm or 25 cm. This study that I’m involved in is one of the first of this kind, looking at the development of a framework for the automated classification of all features within a river. Our initial testing proved to be accurate at 81% and enabled the identification of features at a higher level of detail than that provided by the ground truth data. For example, it allowed the separation of trees from the grass or water underneath, as well as the recognition of hydraulic units (e.g. a rippled water surface).
It’s obvious that we need to protect our rivers and waterways for the benefit of landowners, local authorities and the general population. Better ways of understanding those waterways are just a part of the process – but an important part.
If you’d like to find out more about the use of new technologies for environmental management, you might be interested in my two week UAVs for Environmental Monitoring course, or the Cranfield MSc in Environmental Risk Management.