Global pyrogeography: the current and future distribution of wildfire

Abstract

Climate change is expected to alter the geographic distribution of wildfire, a complex abiotic process that responds to a variety of spatial and environmental gradients. How future climate change may alter global wildfire activity, however, is still largely unknown. As a first step to quantifying potential change in global wildfire, we present a multivariate quantification of environmental drivers for the observed, current distribution of vegetation fires using statistical models of the relationship between fire activity and resources to burn, climate conditions, human influence, and lightning flash rates at a coarse spatiotemporal resolution (100 km, over one decade). We then demonstrate how these statistical models can be used to project future changes in global fire patterns, highlighting regional hotspots of change in fire probabilities under future climate conditions as simulated by a global climate model. Based on current conditions, our results illustrate how the availability of resources to burn and climate conditions conducive to combustion jointly determine why some parts of the world are fire-prone and others are fire-free. In contrast to any expectation that global warming should necessarily result in more fire, we find that regional increases in fire probabilities may be counter-balanced by decreases at other locations, due to the interplay of temperature and precipitation variables. Despite this net balance, our models predict substantial invasion and retreat of fire across large portions of the globe. These changes could have important effects on terrestrial ecosystems since alteration in fire activity may occur quite rapidly, generating ever more complex environmental challenges for species dispersing and adjusting to new climate conditions. Our findings highlight the potential for widespread impacts of climate change on wildfire, suggesting severely altered fire regimes and the need for more explicit inclusion of fire in research on global vegetation-climate change dynamics and conservation planning.

Figure 1. The observed and modeled distribution of fire under current conditions.

(A) Cumulative counts of fire activity detected by the Along Track Scanning Radiometer (ATSR) around the world at a resolution of 100 km over 10 years. (B) The same fire data classified to represent fire-prone (orange) and fire-free (yellow) parts of the world; note that areas of white within terrestrial boundaries were clipped from the analyses to match climate data. (C) Mean of normalized relative probability of fire (nP_c) for ten FIRE_NPP sub-models of fire-prone parts of the world under current conditions. (D) Mean of normalized relative probability of fire (nP_c) for ten FIRE_noNPP sub-models of fire-prone parts of the world under current conditions.

Figure 2. Changes in the global distribution of fire-prone pixels under the A2 (mid-high) emissions scenario.

An increase from current conditions (red) is indicated by a P_Δ greater than unity, little or no change (yellow) is indicated by a P_Δ around unit, and a decrease (green) is indicated by a P_Δ less than unity. Panels show the mean P_Δ for the ensemble of ten FIRE_NPP (A–C) and FIRE_noNPP (D–F) sub-models. Climate projections include 2010–2039 (A, D), 2040–2069 (B, E) and 2070–2099 (C, F).

Figure 3. Potential invasion and retreat of fire.

The invasion (orange) and retreat (blue) of fire projected by 2010–2039 under the A2 (mid-high) emissions scenario and based on the FIRE_NPP ensembles. Invasion was constrained to places with existing vegetation.