The objective of this study was to investigate the response of captive bred and raised koalas to their re-introduction to the wild, particularly with respect to factors operating in the short and medium term following release. Besides studying individual survival, home range and habitat preferences were determined. The preferred habitat selected by the study koalas was used as the basis for modelling habitat suitability outside the primary study area (Brisbane Forest Park). In addition, the predicted suitable habitat distribution was combined with land use data in a GIS platform to perform a "Threshold" approach analysis of threat to the habitat for koala conservation and habitat preservation in the Shire of Pine Rivers (situated adjacent the primary study area).
Radio tracking techniques were employed to monitor the response of the koalas, especially to study their movement in relation to habitat use. ]enrich-Turner home ranges were calculated and plotted on a habitat map in order to depict the general pattern of movement and likely habitat use. The relationship between the movement patterns and the habitat distribution were statistically tested for preference for, or avoidance of, a particular habitat type. A preferred habitat was then identified. This was a vegetation association regarded as a prime habitat focus which was then used as the basis for the prediction of the habitat suitability and availability in the wider study area using a GIS technique.
Thirteen captive bred and raised koalas were released into Brisbane Forest Park where there was once a substantial koala population, which disappeared for unknown reasons in the 1930s. A two year field study was undertaken to observe the movement of the released koalas, and the majority of the captive bred animals became successfully established in home ranges of 16 - 200 ha. The large variation in home range sizes may indicate the variation of the habitat condition leading to optimal selection of the most suitable habitat as social factors were unlikely to have constrained the movement patterns. A statistical analysis of soil nutrients of the home ranges did not reveal statistically significant differences between utilised and non utilised areas, but the data showed a trend for sodium content to be consistently higher in the utilised areas.
Analysis of habitat utilisation was undertaken by overlaying the koala location data and vegetation data of the study area. This area consists of forested habitat of 6 major vegetation types, as well as cleared areas and water bodies. The koalas showed preference for a particular vegetation type, namely open forest with a height of 20-30m "dry sclerophyll forest" having an understory of: grasses, sclerophyllous shrubs, broadleaved shrubs, climbers and ferns. This vegetation type occured on southerly aspects with moderate rainfall or on northerly aspects with high rainfall. The common tree species of this association were Eucalyptus acmenoides, E. decepta (=drepanophylla), E. microcorys, E. propinqua, and E. punctata. An available data set from previous releases of rehabilitated koalas in the same general area was used to test the hypothesis that koalas selected the particular habitat type predicted (or avoided the others). Those koalas showed a consistent selection for the same predicted type.
For the modelling of suitable koala habitat, digital spatial data sets were identified and prepared. These data sets consisted of ARC/NFO coverages of aspect and slope (derived from a digital terrain model), altitude, rainfall and proximity to creeks. These were shown to be the optimal variables in influencing the occurrence and distribution of suitable koala habitat in the current study. Thus, these variables were included for modelling habitat suitability in the wider study area.
The modelling procedure employed two statistical methods: Bayes Theorem and Logistic Multivariate Analysis (categorical modelling) undertaken using the SAS package. Bayes Theorem provides a priori probabilities of the occurrence of suitable koala habitat in a landscape which possesses a set of environmental attributes and combinations; whereas the Logistic Model provides a posteriori probabilities. Reapplication of Bayes Theorem was then undertaken, using as the new probabilities for suitable koala habitat those calculated based on the above a priori and a posteriori probabilities. This set of probabilities was then joined to the spatial data base using GIS to perform spatial analysis of the habitat suitability outside the study area.
At an optimal cut-off point of probabilities applicable in the models, both modelling procedures (Bayesian and Logistic Regression) were able to classify correctly more than 80% of the prime habitat whilst including about 50% of secondary habitat. The results also showed that a spatial model using continuous polygon samples was found to be appropriate for these purposes.
When the extended model for suitable koala habitat was applied in the Shire of Pine Rivers, it was found that the model mainly suited the hilly western part of the Shire (north western part of the Brisbane Forest Park). The eastern coastal plain of the Shire was predicted to be mostly secondary habitat, but koala sighting were common in this area. This limitation of the applicability of the original model was due to the fact that the most significant variable of the model was relief (slope and aspect) which led to the difficulty of the application of the model in the relatively flat area of the eastern part of the Shire. Koala sightings records were used in conjunction with supplementary criteria mainly based on geological information. These have been invoked in order to provide a more generally applicable model in the eastern part of the Shire where the environment substantially differs from the western part and the original study area.
When the model was combined with data on land use in the Shire of Pine Rivers it revealed habitats which were likely to be secure and those which were under threat and needed immediate actions to ensure the protection of their resident koala populations.