Predicting bird habitat resources in temperate woodlands using remotely sensed data : case study in the Great Western Woodlands, southwestern Australia

Abstract

Scant data exist that can inform conservation planners in the Great Western Woodlands (GWW). Yet this extensive area of temperate woodland in southwestern Western Australia is vital for the conservation of many woodland bird species. In this thesis, a modelling approach was developed for predicting the potential distribution of woodland bird habitat functional groups in GWW as a function of their habitat resource use. This model was generated from remotely sensed data-based inputs in conjunction with published bird habitat resource information. Habitat resource information was collected from published literature related to woodland and/or GWW bird species. Based on this information, vegetation structure-related habitat resources were identified and 104 GWW bird species were classified into Bird habitat functional groups (BHFGs). To analyse vegetation structure within GWW, four different satellite-borne data sets (MODIS, ASTER, PALSAR and GLAS) were analysed. The ability of different remote sensing technologies to identify vegetation structure variables was evaluated by comparing remotely sensed data to data collected in the field for selected vegetation sites. The variables were used to model vegetation structure in the landscape of GWW. A bird habitat resource classification model was developed with the nine BHFGs based on the vegetation structure-related habitat resources identified through the literature review. Each spatial data layer derived from the four satellite data sets provided values correlated with five specific Vegetation structure variables: Vegetation cover from ASTER; Foliage density from MODIS; Shrub layer complexity from ASTER; Vegetation volume from PALSAR; and Vegetation height from GLAS. The vegetation structure-based data layers were combined into a three-dimensional Landscape prediction of vegetation structure variables (LPVSV) for predicting bird habitat types derived in terms of the Vegetation structure variables. The BHFGs and the LPVSV were then combined into a Potential bird habitat functional group spatial prediction system (PBHFG-SPS). Due to the lack of field observation on birds in GWW, the predictive capability of the PBHFG-SPS could not be tested. Difficulties in investigating the GWW landscape and the limitations of available data and information about GWW are discussed, as are methods for improving the model developed in this study. The approach developed in this thesis was considered useful for investigating vegetation structure for the purpose of bird conservation, given the limited biophysical field data over extensive and remote areas such as GWW. Show more