Biodiversity responds to increasing climatic extremes in a biome-specific manner.

An unprecedented rate of global environmental change is predicted for the next century. The response to this change by ecosystems around the world is highly uncertain. To address this uncertainty, it is critical to understand the potential drivers and mechanisms of change in order to develop more reliable predictions. Australia's Long Term Ecological Research Network (LTERN) has brought together some of the longest running (10-60years) continuous environmental monitoring programs in the southern hemisphere. Here, we compare climatic variables recorded at five LTERN plot network sites during their period of operation and place them into the context of long-term climatic trends. Then, using our unique Australian long-term datasets (total 117 survey years across four biomes), we synthesize results from a series of case studies to test two hypotheses: 1) extreme weather events for each plot network have increased over the last decade, and; 2) trends in biodiversity will be associated with recent climate change, either directly or indirectly through climate-mediated disturbance (wildfire) responses. We examined the biodiversity responses to environmental change for evidence of non-linear behavior. In line with hypothesis 1), an increase in extreme climate events occurred within the last decade for each plot network. For hypothesis 2), climate, wildfire, or both were correlated with biodiversity responses at each plot network, but there was no evidence of non-linear change. However, the influence of climate or fire was context-specific. Biodiversity responded to recent climate change either directly or indirectly as a consequence of changes in fire regimes or climate-mediated fire responses. A national long-term monitoring framework allowed us to find contrasting species abundance or community responses to climate and disturbance across four of the major biomes of Australia, highlighting the need to establish and resource long-term monitoring programs across representative ecosystem types, which are likely to show context-specific responses.