New remote-sensing system maps Antarctica’s unique vegetation in fine detail

QUT researchers have developed an advanced remote sensing method for accurately detecting and mapping Antarctica’s delicate moss and lichen growth, the mainstays of the continent’s fragile ecosystems. The research team also developed a way to survey Antarctica’s vegetation that is noninvasive and will enable accurate surveys more quickly and cheaply than before. The paper is published in the journal Scientific Reports.

First author and research fellow Dr. Juan Sandino from QUT’s School of Electrical Engineering & Robotics described mosses and lichens as the green “stress barometers” of Antarctica.

“Frost-tolerant vegetation like mosses and lichens in Antarctica are vital to biogeochemical cycles, soil insulation and support of biodiversity,” Dr. Sandino said.

“They drive nutrient cycles and underpin Antarctica’s ecosystems, yet they are the first to suffer from warming, extreme weather and human trampling. Keeping track of their health is vital but extremely difficult in sub‑zero field conditions.”

Dr. Sandino said the researchers flew a UAV (Uncrewed Aerial Vehicle)-hyperspectral camera, which records hundreds of colors for every pixel, combined with Global Navigation Satellite System Real-Time Kinematic (GNSS-RTK) to precisely anchor every pixel to its exact location.

High-resolution RGB UAV imagery was also captured to provide a familiar visual context.

“These three data streams were fused into a streamlined workflow ensuring no moss beds were disturbed,” he said. “This research validated the six proposed spectral indices designed for polar plants which we presented in our previous paper.

“We trained models using these indices which outperformed legacy metrics and found they climbed to the top of every feature-importance chart.

“This new integrated system surpasses conventional digital images (red-green-blue or RGB) and also the satellite-based Normalized Different Vegetation Index (NDVI) that is being used to assess vegetation health and density.”

The researchers compared 12 different AI models for labeling the vegetation and the best options reached about 99% accuracy while staying consistent in rigorous tests.

“This gives us the confidence they will work with future data.”

Professor Felipe Gonzalez, also from QUT’s School of Electrical Engineering and Robotics, said test flights at 30 meters and 70 meters altitude showed that higher flights expanded the mapped area for regional overviews while lower flights captured fine details enabling their approach to scale smoothly from local plots to whole valleys.

“This work proves that a lightweight version using only eight key wavelengths will generate reliable maps resulting in faster more cost-effective vegetations surveys, opening the door for smaller UAVs, lower-cost sensors, and smaller hyperspectral data,” Professor Gonzalez said.