Airborne videographic remote sensing is a relatively recent technology that can provide inexpensive and high-spatial-resolution imagery for forest management. This paper presents a methodology that allows videographic data to be modelled to predict habitat complexity in eucalypt forests. Within the eucalypt forests of south-eastern New South Wales, plots were located on the imagery, and the local variance of the videography within each plot was computed on the assumption that changes in local variance provided an indication of forest structure, and thus the habitat complexity of the site. The near-infrared (NIR) channel demonstrated the most variation, as that channel provided an indication of photosynthetic activity and, as a result, the variation between canopy, understorey, ground cover, soil and shadow provided a highly variable response in the video imagery. Habitat-complexity scores were used to record forest structure, and the relationship between the NIP, variance and field habitat-complexity scores was highly significant (P < 0.001) (r(2) = 0.75; n = 29). From this relationship, maps of the habitat-complexity scores were predicted from the videography at 2-m spatial resolution. The model was extrapolated across a 1 x 1 km subset of the video data and field verification showed that the predicted scores corresponded closely with the field scores. Studies have demonstrated the relationship between habitat-complexity scores and the distribution and abundance of different mammalian fauna. This method allows predictions of habitat-complexity scores to be spatially extrapolated and used to stratify the landscape into regions for both the modelling of faunal habitat and to predict the composition, distribution and abundance of some faunal groups across the landscape. Ultimately, the management of forest habitats for wildlife will depend on the availability of accurate maps of the diversity and extent of habitats over large areas and/or in difficult terrain.
