The innovative use of airborne hyperspectral imaging marks a significant advancement in the quest to identify koala habitats—an issue of critical importance to these beloved creatures. This ambitious initiative, known as 'Project Airbear,' has harnessed advanced technology to map out the types of eucalyptus trees that koalas prefer for nourishment.
Researchers have mounted a cutting-edge hyperspectral imager on a light aircraft, enabling them to scan the landscapes around Gunnedah in New South Wales. This technology works by utilizing narrowband visible and infrared wavelengths to detect subtle characteristics in the trees that are essential for koalas, including leaf pigment, moisture levels, and crucially, nitrogen content.
Koalas are notoriously selective eaters, opting for leaves from specific eucalyptus species that boast the appropriate nitrogen levels. As habitat destruction is one of the leading factors contributing to the decline of koala populations, locating and safeguarding these essential food sources is vital for their survival.
Professor Mathew Crowther, who leads the research at the School of Life and Environmental Sciences at the University of Sydney, likens the search for ideal koala habitats to a 'Goldilocks' scenario—it's not solely about identifying the correct tree species; the nutritional quality must also be conducive to supporting a healthy koala population over time. Remarkably, even within the same species, the nutritional content can vary significantly.
To our knowledge, this is the first study focusing on koalas that employs this method to classify individual eucalyptus species while incorporating tree species into predictions related to nitrogen content. This groundbreaking approach is pivotal as it enhances the speed and efficiency of identifying and protecting critical habitats.
When comparing various remote sensing methods for studying koala habitats, the research published in the 'Science of the Total Environment' reveals a spectrum of techniques. These range from assessing vegetation productivity in relation to koala presence to monitoring habitat health and understanding the impact of bushfire damage on these environments.
Previous studies have attempted to utilize remote sensing to forecast the chemical properties of tree canopies but often faced trade-offs between accuracy and spatial coverage. For instance, while UAVs (Unmanned Aerial Vehicles) can deliver excellent spectral and spatial resolution, they can only cover limited areas. In contrast, satellite imagery can encompass vast regions but lacks the fine detail necessary for precise analysis.
The authors of the current study argue that airborne hyperspectral data strikes a balance between these extremes, making it particularly suitable for regional ecological applications. Unlike satellite multispectral sensors, which offer broader spectral coverage, airborne hyperspectral instruments provide hundreds of contiguous narrow spectral bands. This capability allows for the detection of subtle spectral differences that are vital for predicting leaf chemistry and distinguishing closely related eucalyptus species. Furthermore, the higher spatial resolution helps to reduce mixed-pixel effects, facilitating accurate identification of individual tree canopies, which is especially beneficial for assessing habitat quality.
This research was a collaborative effort involving the University of Sydney, the Sydney Institute of Agriculture, the University of New England, and HyVista Corporation. The study utilized a HyMap hyperspectral scanner, developed by Integrated Spectronics, an Australian firm. The HyMap scanner operates across 128 bands in the reflective solar wavelength range of 0.45 to 2.5 micrometers, offering continuous spectral coverage except in water vapor bands, with an average bandwidth of 15 nanometers. Additionally, its three-axis gyro-stabilized platform minimizes motion distortion, and it undergoes spectral and radiometric calibration using NIST traceable sources. This results in low-noise imagery that is ready for analysis.
In their findings, the authors emphasize the potential of hyperspectral airborne imagery in koala conservation through the identification of premium quality habitats. They highlight the benefits of pixel-based datasets when training models to predict tree characteristics. Despite the inherent noise and variability in pixel-based data—arising from differences in branches, angles to sensors, variations in shading, or background interference—this data allows models to discern real patterns amidst the noise, enhancing prediction accuracy even with limited sample replication, a common challenge in ecological studies.
While average canopy reflectance is still valuable, it should be considered alongside the results derived from pixel-based canopy reflectance. Importantly, the research demonstrates that training models at the pixel level significantly improves their ability to generalize nitrogen predictions, and that explicitly factoring in tree species further boosts model performance. Together, these advancements create a scalable framework for mapping the quality of koala habitats.
Looking ahead, the development of robust remote sensing techniques capable of identifying species composition down to individual plants across landscapes represents a major leap forward for both Australian ecology and Earth observation, according to Professor Bradley Evans from the University of New England. With the enhanced capabilities of the HyVista HyMap sensor, researchers can cover substantially larger areas in a single day, flying at speeds that surpass those achievable with drones.
In the future, we can anticipate more comprehensive assessments of biodiversity and the condition of koala habitats, facilitated by routine flights over Australia’s most critical and vulnerable landscapes. Professor Evans notes that the next phase of the research will involve NASA's Jet Propulsion Laboratory utilizing its latest hyperspectral imager next year. Collaborating with various environmental and agricultural agencies, they plan to have both NASA and HyVista survey the most critical sites.
Furthermore, the University of New England is partnering with the University of Sydney and others to establish a new National Collaborative Facility dedicated to this technology, ensuring that these advancements are effectively employed for the benefit of Australia and future initiatives with NASA.
