Habitat mapping and species distribution modelling of the endangered Bulloak Jewel Butterfly

Habitat mapping and species distribution modelling of the endangered Bulloak Jewel Butterfly

Assessing and mapping of wildlife habitat plays a vital role in the development of management plans and conservation strategies. Habitat assessment commonly involves the estimation of population size, identification of critical habitat and prediction of environmental change impacts. However, mapping of wildlife habitat, as well as modelling its species distribution at moderate to fine scale levels, is not trivial especially for poorly-studied landscapes. The broad aim of this research was to develop a methodological framework in assessing, mapping and modelling the distribution of an endangered species, i.e. the Bulloak Jewel Butterfly (Hypochrysops piceatus).

The habitat preferences for Bulloak Jewel Butterfly are complex due to its mutual relationship with bulloak tree (Allocasuarina luehmannnii) and a little known ant (Anonychomyrma sp.) species. This butterfly is endemic to Australia. Its distribution is very limited and only located west of the Darling Downs (Queensland), near Goondiwindi and Leyburn. Enhanced and incessant study on the restricted distribution of this butterfly is essential. The Bulloak Jewel Butterfly was last sighted in 2003 and requires urgent attention particularly on its habitat management and conservation planning as it is potentially exposed to habitat loss caused by the impacts of land cover changes.

In this research, remote sensing and GIS technologies were applied through the application of hyperspectral data (field sensor), Landsat-5 TM, Landsat-8 and WorldView-2 imagery, vegetation samples, relevant thematic maps (i.e. remnant ecosystem, soil, land use, topography, foliage projective cover, and proximity layers) and species occurrence (i.e. presence-only) data. A suite of analysis techniques, namely: partial least squares (PLS) regression, feature extraction, morphological spatial patterns (MSPA) and connectivity analyses, and maximum entropy (MaxEnt) model, were employed at various components of the study. These analyses were performed according to a specific objective covered by each 'technical' chapter (Chapters 4, 5, 6 and 7).

Spectral analysis of hyperspectral data using PLS regression highlighted significant findings in discriminating seven woodland species that are associated with bulloak tree species, at leaf and canopy levels. The predicted and measured values for the validated samples in PLS were highly correlated with r = 0.985 to 0.997 and r = 0.985 to 0.996 for canopy and leaf levels, respectively. The root mean square errors of prediction (RMSEP) values were reasonably low and indicated high discrimination accuracies which were 89.14 to 95.67% and 91.73 to 94.58% for canopy and leaf levels, respectively. Furthermore, the near infra-red (NIR) region (700-1355 nm) of vegetation reflectance property was found to have a significant role in the discrimination between plant species for both leaf and canopy levels. Thus, this finding signified that the bulloak tree (dominant Bulloak Jewel Butterfly) can be easily recognised using hyperspectral data. In addition, the assessment of habitat attributes or habitat quality of Bulloak Jewel Butterfly were also established using vegetation structure analysis. This analysis incorporated WorldView-2, Landsat-5 TM and Landsat-8 imagery, together with field vegetation samples. The relationship between WorldView-2’s raster layers and vegetation structure attributes was confirmed to have the highest correlation among those three satellite imagery, ranging from r = 0.650 to 0.719 (ρ < 0.05).

Feature extraction analysis on the high resolution imagery (WorldView-2) generated an object-based classification map with high overall accuracy value, i.e. 90.54%. It helps in identifying the vegetated and non-vegetated areas since vegetated area was anticipated to be the Bulloak Jewel Butterfly’s preferred habitat. The performance of Landsat-5 TM and WorldView-2 imagery in assessing the landscape pattern of the Bulloak Jewel Butterfly habitat were done through the execution of MSPA and connectivity analysis. MSPA was significant in mapping and monitoring the spatial pattern of forest (Bulloak Jewel Butterfly’s habitat) versus non-forest areas at the pixel level. WorldView-2 always produced higher percentage of class 'core' (patch or forest area) than Landsat-5 TM in all conditions; either for raw or convolution outputs. These situations indicated that the higher the spatial resolution of the imagery, the higher the 'core' area extent mapped and consequently formed more details of habitat extent.

Furthermore, MaxEnt was a useful model in modelling and mapping the current distribution of Bulloak Jewel Butterfly, as well as in assessing the impact of potential land cover changes towards its habitat suitability. The significant contribution of different predictor variables produced by six MaxEnt models in assessing the current (recent condition) distribution of Bulloak Jewel Butterfly species also highlighted in this analysis. Those models were indicated by the different predictor variables used in each model; Models 1 to 6. The results have shown that Model 6 (all uncorrelated imagery predictors) signified the lowest AUC (area under the ROC curve) value (0.880). However, Models 1 to 5, which have both imagery and feature-based layers as their uncorrelated predictor variables, produced AUC values better than 0.940. Model 4 (highest AUC value of training data per category) was elected to be the preeminent MaxEnt model in this study based on the highest number of ‘high’ suitability area of Bulloak Jewel Butterfly habitat, appropriate number of predictor variables used and high AUC value. Model 4 could also assist in executing one of the actions needed: 'locate additional colonies and clarify and map the present distribution and habitat' as indicated in the Recovery Plan for the Bull Oak Jewel Butterfly (Hypochrysops piceatus).

Additionally, the outputs of scenario models given by Current Model 1 (CM1, with RE and NDVI as the inputs) and Current Model 2 (CM2, with RE, NDVI, slope, landuse, and distance from the water body as the inputs), as well as their corresponding projected models, were visually and statistically diverse. However, the habitat suitability patterns given by Current Models 3 and 4 (CM3, with FPC as the input; CM4, with FPC and RE as the inputs) and their corresponding projected models were marginally the same. This variability occurred due to different set of predictor variables used. Model CM2 and its corresponding projected model were established to be a 'better predictor' of those tested modelsl in predicting the impacts of land cover changes.

In conclusion, this research shows that the established methodological framework is useful in assessing, mapping and modelling the habitat of Bulloak Jewel Butterfly. High spatial resolution imagery of WorldView-2 is much more sensitive and most relevant dataset to be used in assessing and mapping the habitat of Bulloak Jewel Butterfly. Species distribution models developed in this study are also valuable in designing conservation strategies for this butterfly species. The analysis of climate change, particularly changes in rainfall and temperature, and how these attributes could distress the suitability of Bulloak Jewel Butterfly habitat could be a relevant study for future work. The novelty of this research are mainly in filling the research gaps on the requirement for mapping habitat attributes at fine and medium scales, and how satellite imagery and map layers can be used to produce image-derived habitat maps. Furthermore, the significant contribution of this research is on the novel integration of the outputs of feature extraction module, morphological analysis of forest spatial pattern in conjunction with landscape network connectivity and species distribution modelling which could offer substantial and strategic inputs for the conservation strategy of the Bulloak Jewel Butterfly and its habitat. This type of methodological framework is not just limited for the purpose of conservation strategy of Bulloak Jewel Butterfly only, but also significant to other wildlife especially to those who has connection with plant species.