Impacts of climate, CO2 levels and human activity on the spatial distribution of global fire regimes

Abstract

Wildfires are fundamental to the distribution of vegetation on Earth. Wildfires vary in size, frequency, and intensity and are controlled by climate, vegetation, and human activity. A fire regime refers to the long-term, repeated patterns of wildfires under a given set of conditions. Shifts in global fire regimes in response to changing environmental conditions are a growing concern. Many global vegetation models include wildfire but struggle to predict fire properties beyond the first-order patterns of burnt area and disagree on its environmental responses and historical trends. This thesis develops robust empirical models of burnt area, fire size and a measure of intensity and uses them to explore the global sensitivity of fire regimes to changes in climate, atmospheric CO2 and human activity. Simulation of fire regimes under the very different conditions of the Last Glacial Maximum predicts reduced fire activity consistent with evidence from sedimentary charcoal records, while allowing for the separation of CO2 and climate change effects (both significant). Simulation of future fire regimes under low- and high-mitigation scenarios indicates a global shift in wildfire patterns by 2100 CE with burning reduced in tropical regions but larger and more intense wildfires in extra-tropical regions. Under low mitigation, increases in burnt area worldwide overwhelm the current human-driven declining trend, with fire size and intensity increasingly limited by dryness and vegetation fragmentation. These model experiments highlight the distinct controls of different fire properties. Whilst burnt area is driven by fuel availability and dryness, fire intensity is limited by fuel build-up, and fire size by fuel continuity. This decoupling occurs because of the different temporal and spatial scales on which the controls 6 of burnt area, fire size and fire intensity operate. These findings have immediate implications for the improvement of process-based fire models, which currently do not take these distinctions into account.