Fusing spectral and structural imagery detects escalating canopy dieback in a vulnerable eucalyptus population

Pixel‑based classification of canopy condition into healthy, moderately stressed, and stressed categories achieved high accuracy (Kappa = 0.81) across both surveys. Results showed a decline of over 50% in the healthy canopy class. Overall, 25.95% of the canopy displayed a positive change and 17.71% a negative change between the two surveys.

Structural and spectral indicators indicated different patterns. Normalised LiDAR return density increased, while NDVI decreased in previously healthy sites and increased in unhealthy sites. This pattern likely reflects resprouting of drought-stressed trees and proliferation of understorey vegetation during wetter years, while NDVI decreases in previously healthy sites likely reflect canopy browning under drought, suggesting growing vulnerability for the study population under a warming, drying climate.

More broadly, our study highlights the value of integrating spectral and structural data to detect, monitor, and understand the intricacies of changes in canopy health. Hence, integrated spectral and structural data may allow determination of vegetation change and landscape vulnerability under climate change at large scales. Identifying these changes and vulnerabilities is increasingly important, as the speed and extent of climate change impacts make field-based detection increasingly challenging.</p>