Detecting a fire-sensitive species in a fire-prone landscape: object-based rule-set driven approaches

The Northern Cypress Pine, Callitris intratropica (fam: Cupressaceae) is a fire sensitive conifer that exists in a range of environments across the tropical savanna landscapes of Australia's Northern Territory. Its susceptibility to intense fires means that populations are vulnerable under late season high intensity burning regimes, limiting the range to areas that are protected from such fires. The presence or absence of C. intratropica is seen as an indicator of the fire regime within a landscape. Methods of detecting individuals and stands of the species from high resolution remotely sensed imagery will assist in assessing the distribution of the species within the landscape. The work here describes two object-based methods: one region growing and the other a thresholding technique, for detecting stands and individuals of the species from high spatial resolution multispectral satellite imagery from within Eucalypt dominant savanna. The region growing algorithm identifies 'seed objects' within trees based on maximum Normalised Difference Vegetation Index (NDVI) values and then grows these objects to encompass the tree crowns. The threshold segmentation technique creates objects of woody cover in the savanna using a NDVI threshold. C. intratropica is then differentiated from other woody cover using thresholds based on the near infrared and red bands of the data. Detected crowns and stands are then compared against recorded field and manually delineated observations. Based on the reference data, accuracies for both methods is over 70% showing the potential for mapping the species across a wider area.