Impacts of habitat and climate change on the abundance, distribution, and functional connectivity of pollinators in Australia

Pollinators fulfil a crucial service in terrestrial ecosystems, by pollinating crops and wild plants. Yet they are threatened by the synergistic effect of several factors, including human land-use and land-cover change, and global climate change. The impact of those threats can be mitigated by conserving and promoting functional landscape connectivity, which is essential for maintaining ecological processes such as pollination. However, pollinator-landscape studies typically focus on a few focal taxa and areas undergoing substantial change. Further, the assessment of the impacts of climate change, extreme-temperature events, and land-use and land-cover change, on the distribution and functional connectivity of pollinators, is limited, with pollinating invertebrates being particularly poorly studied in this context. In this thesis, I sought to fill some of these knowledge gaps. Firstly, I determined the association between land use, land cover, plant genera, and pollinator abundance (bees, beetles, and nectarivorous birds) in a mixed-use landscape in southern Australia, and determined the most effective ways of using remote-sensed data to quantity land cover over larger spatial scales. I then predicted the combined impacts of climate change and extreme-temperature events on the future distribution of fruit bats in Australia. Here, I used fruit bats as a case study, because compared to other pollinators, they are a subject of human-wildlife conflicts, such that range shifts by fruit bats can have negative implications. Finally, I assessed the landscape functional connectivity for butterflies between protected areas under different scenarios of land-use and land-cover change and climate change in Australia. I used butterflies as a case study because existing studies on functional connectivity are biased towards birds and mammals, and among different invertebrate groups, butterflies are better studied, and so can be adequately modelled in ways that can address this key issue. Using mixed-effect models, in Chapter 2, I demonstrated that protected areas did not support a higher abundance of pollinators than plantations and pasture land use. Protected areas and plantations had a positive effect on honeyeater abundance while pasture had a positive effect on introduced bee abundance. Forest within 100 m buffer had a positive effect on honeyeater abundance while open land cover within 250 m and 1500 m buffers had a positive and negative effect on native and introduced bee and honeyeaters respectively. The native plant genera Acacia, Leptospermum, Leucopogon, Lissanthe, Melaleuca, Pimelea, Pomaderris, Pultenaea, had a positive effect on the abundance of native and introduced bees and beetles while the subgenus Symphyomyrtus had a positive effect on honeyeater abundance. In Chapter 3, which is an extension of the methodology of Chapter 2, I showed that using an ensemble of statistical-learning algorithms is an effective way of inferring land cover from satellite imagery and can overcome difficulties in model selection. Between using an unweighted and weighted ensemble, both achieved similar accuracies, but the unweighted ensemble is a simpler option of the two given the number of ways weights can be calculated. In Chapter 4, using correlative species distribution modelling, I showed that the distribution of six species of fruit bats in Australia is predicted to remain stable in the coming decade. However, all species are still predicted to lose and gain areas and these changes are predicted to occur along the edges of the species' current distribution range. Of all our study species, only the Grey-headed flying fox (Pteropus poliocephalus) is predicted to have high suitability, under present conditions and future scenarios, across southwestern Australia and the island of Tasmania, even though these areas lack fruit bats. We speculate that it might be possible for fruit bats to occupy these areas in the future given their ability to travel large distances and colonise previously uninhabited areas. However, whether they will be driven to occupy these areas permanently, directly or indirectly because of climate change and extreme temperature events, is yet to be determined. Using connectivity analysis (Circuitscape) in Chapter 5, I found functional connectivity models are predicted to decrease from a cumulative current (a proxy for species‚ÄövÑv¥ movement) of 0.0191 under present conditions to 0.0163 and 0.0162 under future scenarios 2050 and 2090 respectively. Cumulative current is predicted to decrease and increase for 30 and 20 species, respectively, while 9 species are predicted to mixed trends depending on the future scenario. Overall, these changes are generally predicted to occur along the edges of a species‚ÄövÑv¥ current distribution. In conclusion, difficulties in algorithm selection can be overcome using ensemble algorithms which can be adopted for land-cover classification and to model habitat suitability. Importantly, the case studies presented in this thesis showed that land use, land cover, and climate change have varying predictive capacity and effects on pollinators depending on the group or species studied, suggesting no one size fits all. Yet, some results were common across taxa. For instance, future changes in both distributional range (for fruit bats) and functional connectivity (modelled for butterflies) will likely occur mostly along the periphery of a species‚ÄövÑv¥ current distribution, pointing at areas within the core of a species range to be key locations to support species survival under climate change. Once those sites are identified, specific information on pollinator habitat preferences, and particularly which plant species best support the taxon of interest (or multiple taxa), can inform local management strategies.