Researchers have utilized airborne hyperspectral imaging to map the distribution of eucalyptus trees preferred by koalas. This initiative, named 'Project Airbear,' involved scientists using a sophisticated hyperspectral imager on a light aircraft to scan areas near Gunnedah, NSW.
Technology and Purpose
The technology employs narrowband visible and infrared light to identify specific characteristics within trees, such as leaf pigment, water content, and nitrogen levels. Koalas are selective eaters, preferring particular eucalyptus types with appropriate nitrogen content. Given that habitat loss significantly contributes to koala population decline, identifying and conserving suitable food sources is essential.
Professor Mathew Crowther from the University of Sydney stated that finding optimal koala habitats is a 'Goldilocks' situation, requiring not just the correct tree species but also the right nutritional quality. This study is noted as the first koala-focused effort to classify individual eucalyptus species using this method, including tree species as a factor to improve nitrogen content predictions. The technology is expected to accelerate habitat identification and protection.
Remote Sensing Comparison
The study, published in Science of the Total Environment, reviews various remote sensing techniques used for koala habitat research. While UAVs offer high resolution for small areas and satellites cover large areas with coarser resolution, airborne hyperspectral data provides a balance, making it suitable for regional ecological applications. Airborne hyperspectral instruments offer numerous narrow spectral bands, enabling detection of subtle differences crucial for predicting foliage chemistry and distinguishing related species. Their finer spatial resolution also helps minimize mixed-pixel effects, allowing accurate individual tree canopy delimitation for habitat quality assessments.
Study Details and Findings
The research was a collaboration between the University of Sydney, the Sydney Institute of Agriculture, the University of New England, and HyVista Corporation. HyVista utilized an Australian-manufactured HyMap hyperspectral scanner.
The research indicates the potential of hyperspectral airborne imagery for koala conservation through identifying high-quality koala habitat. Findings emphasize the value of 'pixel-based' datasets for training models to predict tree features. This approach allows models to capture patterns despite data noisiness, improve predictions, and remain effective even with low replication rates common in ecological studies. The study demonstrated that pixel-level training significantly enhances model generalization for nitrogen prediction, and incorporating tree species further improves performance, providing a scalable framework for mapping koala habitat quality.
Future Plans
Professor Bradley Evans from the University of New England highlighted that robust remote sensing techniques identifying species composition down to individual plants are a significant advancement for Australian ecology. The next phase of research will involve NASA JPL using its latest hyperspectral imager. Collaborations with environmental and agricultural agencies are planned to fly critical sites with NASA and HyVista. The University of New England, in partnership with the University of Sydney and others, is planning a new National Collaborative Facility for this technology.