Essay on Earth and Environment Communications

Daniel Paes, a PhD student in Geology at the University of Sciences in Lisboa, is one of the authors of the article – “Benchmarking satellite-derived coastal mapping algorithms“ – published in the journal Communications Earth & Environment, last September provides an unprecedented assessment of the accuracy of shoreline detection, using publicly available satellite images.

Satellite remote sensing, due to the abundance and accessibility of data, is often used for coastal monitoring, but until this article, there has been no systematic and independent comparison of the performance of different algorithms that, from satellite images, can automatically detect He drew a map of the coastline, that is, the point of contact between land and sea.

As part of his PhD thesis, Daniel Paes was invited by his external co-supervisor & Luis Pedro Almeida – Researcher at the Portuguese Collaborative Laboratory Colab + Atlanticto participate in this innovative project he designed Kilian Voss, first author of the article and a researcher at the University of New South Wales (Australia). This work also includes the participation of many experts from the following: Polytechnic University of Valencia (Spain), University of Queensland (Australia), Deltares Institute (Netherlands), Federal University of Santa Catarina (Brazil), University of Bordeaux (France). USGS Coastal and Marine Science Center.

Luis Pedro Almeida and Daniel Paes and Antonio Klein (UFSC) tested the tool Cassie (short for Space Imagery Engine Coastal Analysis System), launched Connected In 2021, in the framework of a Portuguese-Brazilian collaboration still under development, which automatically detects and analyzes the coastline using satellite images. According to the student, “CASSIE presents a user interface Intuitive graphics, with visual indicators, which makes it different from other, more complex indicators out there. Anyone who is not a specialist in the area can use the tool and understand what is happening in a particular coastal area.

The practical work lasted about six months. The first objective focused on operating the CASSIE tool in coastal control environments: four sandy beaches with varying geological and oceanographic conditions. Due to the dynamic nature associated with the location of the coastline, there was a need to correct the results, incorporating tidal and wave-related variations, in order to increase detection accuracy. This aspect also revealed a high dynamic component to the indicator used, especially in more complex coastal environments.

“In places where the tidal range is larger, algorithms tend to give worse results because they cannot detect the coastline correctly. One of the conclusions of this article is that in these conditions it is necessary to develop new specific techniques and even, perhaps, use a different indicator such as the limits of Dry sand/wet sand “In areas with smaller tidal ranges, the results are very positive,” explains Daniel Paes.

Another important part of the work is validating the products generated by automatic shoreline detection algorithms using data in the site collected in the coastal environments under analysis. “This was only possible due to the fact that these four sites have been monitored for many decades, very frequently, and this data is publicly available, which is relatively rare in the world.” Data collection On site“In the field, it will always be essential and necessary to validate any model,” explains Daniel Paes.

In his doctoral project, and at a more advanced stage, Daniel Paes intends to explore coastline prediction, adding other factors such as waves, tides or sediment supply to satellite-detected lines to create scenarios that represent their future development and make them useful to society, making the results available to entities. Interested public.

This project opens doors to the development of new algorithms, new technologies and a greater understanding of coastal environments on a global scale, and there is also scope to improve the tested tools themselves, such as CASSIE.

“The methodology followed, completely transparent and collaborative, allows the work to be replicated by other researchers. This is something important in science and we wanted to convey it through this project,” concludes Daniel Paes. The next step is to improve detection through artificial intelligence, incorporating neural networks into satellite image analysis, and testing this method with high-resolution (3m) satellites and in other coastal environments.