AI-based oxbow lakes
recognition system

New technologies based on computer vision and advanced data analysis are increasingly being used in environmental and nature conservation efforts. As part of the "Most Valuable Rivers and Streams in Poland" initiative, experts from WWF Poland, the University of Life Sciences in Poznan, and deepsense.ai joined forces to create an intelligent system to recognize oxbows, an essential element in the protection of Poland's river ecosystems. This unique project is the first attempt to identify all oxbows in Poland using advanced AI technology.

deepsense.ai developed a method for the automatic identification of oxbows based on data pertaining to their shape and location. These shallow water bodies are remnants of a former river bed and essential to the preservation of the river's natural ecosystem. Oxbows support numerous rare and protected species of flora and fauna and play a significant role in water retention and river purification. Identifying and monitoring oxbows is thus crucial to effective conservation efforts. Identifying oxbows manually would be extremely challenging due to the massive number of bodies of water, making it necessary to examine each individual body on maps and satellite images.

Instead of manually identifying oxbows, the system developed by deepsense.ai efficiently analyzed a large data set and distinguished oxbows from other natural and artificial water bodies. The process, which would typically require a team to undertake many months of intense work, was reduced to a few hours thanks to the implemented AI solution.

Source: Google Earth

Source: Screenshot from the system

The development of the oxbow identification system was a multi-stage process that lasted over six months. The AI model was created using data from about 2,000 small water bodies in river valleys and was able to categorize water bodies into one of three categories: oxbows, natural non-oxbow water bodies, or artificial water bodies. The model considered various aspects, such as the shape of the analyzed object, its size, and the distance from the riverbed. Initial results from the AI model indicated that there are approximately 22,000 oxbows in Poland. To test the AI system's effectiveness and refine the model, researchers manually identified a random sample of 1,000 water bodies. This meticulous work was necessary to achieve a high level of efficiency in the identification of oxbows.

The AI-based recognition system will be used for comprehensive assessments of the natural state of river valleys to identify the most valuable rivers and valleys requiring protection, such as nature reserves. This evaluation will also enable the identification of rivers and sections which are suitable for renaturalization, leading to a significant improvement in degraded river ecosystems. In addition, the identification of oxbows can be used to help increase the resistance of rivers to natural disasters.

The final outcome of the collaboration between environmentalists and deepsense.ai experts will be two AI-powered tools. The first is a model for identifying oxbows in large areas, which can be applied in countries other than Poland, while the second is an educational online application for students and lecturers at higher education institutions that allows for the analysis and classification of up to 100 water bodies based on spatial data entered by the user.