From operational experimentation to scientific research

The Hydrographic Institute, in collaboration with the NOVA Information Management School and the Faculty of Sciences of the University of Lisbon, has published the scientific article “Improving UAV-Derived Nearshore Bathymetry Using Geospatial Machine Learning” in the IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

The study uses experimental data collected during the Portuguese Navy's REPMUS – Robotic Experimentation and Prototyping with Maritime Unmanned Systems exercise to investigate new approaches for determining nearshore shallow water bathymetry, which is traditionally difficult to monitor using conventional hydrographic means.

Combining multispectral imagery acquired by an unmanned aerial vehicle (UAV), multibeam echosounder bathymetric data, and geospatial machine learning techniques, the work assessed the capability to estimate depth from the UAV's multispectral images. The approach, based on Geographical Random Forest, made it possible to explicitly account for the spatial variability of the relationships between reflectance and depth, while simultaneously improving predictive capability and spatial interpretation compared to a conventional Random Forest model.

Beyond the scientific results, this work highlights a relevant dimension of REPMUS: the data obtained during experimentation can continue to generate knowledge long after the exercise period, supporting new methodologies, academic research, and solutions for ocean observation and mapping.

It is also an example of how cooperation between operational and academic institutions allows transforming experimental data into reusable scientific knowledge, strengthening the link between experimentation, innovation, and ocean science.

Access the article here:
https://www.doi.org/10.1109/JSTARS.2026.3721930