How artificial intelligence is helping track beach changes using smartphone photos

The Balearic Islands Coastal Observing and Forecasting System (Socib), working with the Artificial Intelligence Research Institute at the Universitat de les Illes Balears (UIB), has developed an artificial intelligence tool that automatically identifies the shoreline in photographs taken by members of the public on their smartphones.

In a statement issued on Tuesday the 10th of November, the UIB explained that the research addresses the need to monitor beach changes more frequently in response to erosion, storms and rising sea levels.

Photographs contributed by the public allow the shoreline to be mapped at a higher spatial resolution than satellite imagery provides.

The system uses images submitted through CoastSnap, an international initiative that provides fixed smartphone cradles so that photographs are always taken from the same viewpoint.

In the Balearic Islands, Socib maintains stations at s'Amarador beach in Santanyí and s'Arenal d'en Tem beach in Campos. It plans to install ten more over the coming years, with four in Eivissa, three in Mallorca and three in Menorca.

These images reveal changes in the position of the shoreline, an essential indicator for identifying erosion, post-storm recovery and longer-term coastal trends.

To develop the tool, the team trained and compared four deep learning models, a form of artificial intelligence, using 1,717 images from five Atlantic and Mediterranean beaches.

The most effective model, DeepLabV3, analyses the full context of an image, much like a trained eye, rather than simply detecting colour changes. This enables it to identify the shoreline even in the presence of sun glare, shadows, wet sand, vegetation debris or flooded areas.

The results show that the new system is considerably more accurate than conventional automated methods, reducing error by approximately 70 per cent in the study’s overall comparison. Its results also closely match those obtained by specialists working manually, with median differences of just 2.6 metres at s'Arenal d'en Tem and 3.4 metres at s'Amarador.

The study’s lead author and Socib data scientist, Josep Oliver-Sansó, said, “Automating this process allows us to turn thousands of photographs contributed by the public into useful information to understand how beaches are changing and inform adaptation decisions in response to the effects of climate change.”

The software has been released as open source and can be integrated into other coastal monitoring systems. It can process an image in approximately six seconds on a standard computer and in less than a second using a graphics card.

The tool is also available through European open science platforms such as iMagine and AI4EOSC, making it easier to reuse and integrate into other coastal observing systems.

The collaboration between Socib and the UIB has thus brought citizen science closer to public management, enabling more continuous data collection to support coastal protection and adaptation planning.