Spatial and temporal expansion of global wildland fire activity in response to climate change

Present fire-climate classification

To identify the different regions of the planet with suitable climatic conditions for fire activity, we compare the global distribution of climate indicators based on temperature and precipitation, with satellite-derived GFED4 burned area data²¹ (Fig. 1). Starting from four general climates (Tr-tropical, Ar-arid, Te-temperate and Bo-boreal) based on the Köppen–Geiger climate classification main categories²², we create four fire-prone classes using climate thresholds to define the patterns observed in Fig. 1. Each category is characterised by the element that boosts fire activity during the FS: low precipitation, high temperatures or a combination of both. The classification is made by contrasting the probability distribution of the climatic variables at data points associated with high fire activity vs. points with low fire activity within the main Köppen-Geiger categories (see Threshold Selection in Methods section for a detailed explanation).

The environmental conditions associated with fire occurrence emerge more clearly in this comparison, yielding the different threshold sets in Table 1 that determine the fire-prone months at any location (the selection method is detailed in the Methods section). We define those years with at least 1-month meeting the thresholds, as fire-prone years (FPY). Depending on the number of FPY at each location, the categories of Table 1 are sub-divided into recurrent (r), occasional (o) and infrequent (i) (see Methods). The average number of fire-prone months during the FPY is defined as the potential FS length (PFSL), i.e., the season with climatic characteristics prone to fire activity.

Figure 2a depicts the global map of the burned areas classified according to the selected thresholds (Table 1). Savanna fires are responsible for the largest proportion of burned area on the global scale²¹. The FS in these areas is longer than in other climates (see Supplementary Fig. 1) and, despite savanna fires being also dependent on ignition patterns and human policies and practices, the FS is tied to a pronounced seasonal cycle of precipitation^23,24,25, with fire occurring mainly during the dry part of the cycle. Because of this, the Tropical – dry season fire class (Tr-ds) coincides with the distribution of the tropical savanna climate. In Fig. 2, boreal fires are represented as hot season fires (Bo-hs) due to the large positive temperature anomaly existing in those locations during the FS (Fig. 1c). In fact, temperature variations explain much of the variability in boreal burned area^26,27. Temperate fires are classified as dry and hot season (Te-dhs) because they affect regions where the dry season coincides with the warm season (Fig. 1b, c). Here, high temperatures and precipitation seasonality determine fire activity and inter-annual burned area variability, e.g., in Western North America^28,29,30,31 and Southern Europe^32,33. Fire activity in arid regions occurs during warm months, but the relation with precipitation is more complex. The FS is associated with a hot season in cooler (MAT < 18.5 °C) midlatitude arid areas where no clear wet period is observed, e.g., the Western US and Central Asia (Supplementary Figs. 10 and 11), but closer to the tropics where it is warm year-round, it can be also determined by the existence of a marked annual wet and dry season cycle, with fires occurring sometime during the dry season. In the warmest arid regions (MAT > 27.5 °C), the FS starts right at the beginning of the dry season (e.g., the Sahel, Supplementary Fig. 12) while where MATs are more moderate, between 18.5 and 27.5 °C, it takes longer to develop (e.g., Central Australia and the Kalahari desert, Supplementary Figs. 12 and 13). Due to the dependency between fires and the existence of fuel in arid climates, we named this class Arid fuel limited (Ar-fl). A more in-depth discussion about the definition of this fire-climate class can be found in the section entitled Threshold selection for each climate of the Supplementary Information.

In Fig. 2b, we classify every spatial point and not only those with burned area observations as in Fig. 2a. The four groups in Fig. 2a show observed fire locations that share some specific climate conditions, and the classification in Fig. 2b shows the world areas where these conditions occur for at least one month. The correspondence between these two maps is quantified in Supplementary Fig. 20b, with more than 70% of the land area well classified as either fire-prone (BA > 0 ha) or fireless (BA = 0 ha). This reveals a two-way relation between fires and climate: fires take place under specific climatic conditions, and most places with these climatic conditions are indeed fire-prone, which supports our earlier hypothesis. Fire activity is controlled by weather, resources to burn and ignitions, as represented through the fire regime triangle^12,20. On broad temporal scales and large spatial scales, temperature and precipitation have an important impact on fire because these climate variables influence vegetation type and the abundance, composition, moisture content, and structure of fuels³⁴. Although ignitions may be driving fires to a greater extent than temperature or precipitation at specific locations or events³⁵, they do not seem to limit fire activity at coarse spatial and temporal resolutions, implying that where fuels are sufficient and atmospheric conditions are conducive to combustion, the potential for ignition exists, either by lightning or human causes^13,20. For all these reasons, we can identify specific climates that are prone to fires.

The areas classified as fire-prone in Fig. 2b comprise 99.26% of the observed global mean annual burned area in Supplementary Fig. 2. This percentage is above 85% for all four general climates (Supplementary Fig. 20). The percentage of land area with non-zero burned area data classified as fire-prone is 91.22%. Considering for each location only the obtained FPY, the percentage of the observed burned area classified is 90.36%. Furthermore, the PFS obtained in the fire-climate classification (Fig. 3b) also correlates well with the timing of observed fire incidence, as globally 87.91% of the observed mean burned area occurs during the identified months of PFS at classified fire-prone locations.

Unclassified regions (in grey in Fig. 2a) correspond for the most part to those with the least burned area or those where agricultural practices modify the climatic seasonality of fires. In addition, as the classification is conceived from a climatic point of view, locations with fire activity associated with specific meteorological conditions that are not appreciable at the monthly temporal resolution, are probably not well identified. For example, a week of extremely high temperatures could be almost unnoticeable in the monthly mean temperature, but not in fire activity. Similarly, months with the same total precipitation may have different fire activity if the precipitation falls concentrated in a few days or is distributed throughout the month. Furthermore, the short temporal sampling period of the burned area data could also be influencing our results. Locations with long fire cycles may not be well represented in the data.

Future fire-climate classification

A future fire-climate classification map is derived by applying the thresholds obtained in the present fire-climate classification to future climatology variables from multiple coupled model intercomparison project phase 5 (CMIP5) global circulation model (GCM) outputs, considering the RCP8.5 scenario (the worst-case climate change scenario of the CMIP5). Two contrasting approaches can be taken for analysing future fire activity, one that considers quick vegetation adaptation to the new climatic conditions, and another that does not. These two approaches clearly diverge in the boreal regions, where the biome (mainly taiga) is strongly conditioned by the low temperatures and where future temperature changes at the end of the 21st century will have a greater amplitude. It is expected that the boreal forest of these areas will not be immediately replaced by a temperate mixed forest where the average annual temperature exceeds the range of values typical of the taiga biome. Terrestrial vegetation compositional and structural change could occur during the 21st century where vegetation disturbance is accelerated or amplified by human activity, but equilibrium states may not be reached until the 22nd century or beyond³⁶.

Based on the assumption that during the future period (2070–2099) the vegetation will not be fully adapted to the new climatic conditions, and since the present Köppen–Geiger climate classification (on which we base our Tr, Ar, Te and Bo categories) closely corresponds to the different existent biomes²², we analyse only the projected changes in the specific fire-climate classification variables, maintaining the general division of Tropical, Arid, Temperate and Boreal regions as is in present climate conditions. The future fire-climate classification is shown in Fig. 2c.

We note that we determine future fire activity from relationships of the latter with the present climate; however, these relationships might not be stationary. Our approach does not contemplate possible future changes in precipitation frequency if they are not noticeable in monthly precipitation amounts. Areas with the rising incidence of extreme precipitation events due to global warming³⁷ may experience an increase in monthly precipitation but a decrease in rainy days, which may lead us to consider the conditions there less favourable for fire activity than they actually will be.

Future changes in global fire activity

Modelled future fire-prone regions experience significant variations with respect to the present (Fig. 2b, c). Due to global warming, the Bo-hs fire class pertaining to boreal forests would spread over a larger area, reaching most of Northern Scandinavia and undergoing a southward and northward expansion in Canada, Alaska and Russia. This category may experience a percentual expansion of 47.0% according to our results. This expansion is more accentuated for the combination of the highest recurrence subcategories Bo-hs-r and Bo-hs-o, reaching a value of 111.5%.

The conjunction of Te-dhs-r and Te-dhs-o fire classes of midlatitudes also undergoes a considerable expansion of 24.5% in the area (Fig. 2b, c). The most remarkable changes are expected in Southern China and Southern Europe. A large part of Europe transitions from an infrequent fire category to a more frequent fire category with Csa and Csb Mediterranean climates³⁸.

The Tr-ds fire classes with frequent fire-prone conditions in the Tropics presents fewer spatial changes (Fig. 2b, c), with a spatial contraction of 6.3%. The most important differences are found in South America. Some of the climate model results considered here indicate also that some parts of the Eastern Amazon rainforest will move from a non-fire class to Tr-ds fire class, as other studies have suggested³⁹.

The Arid fire-prone classes Ar-fl-r and Ar-fl-o would increase its area by 5.0%. Projected changes in the extent of this class are very sensitive to changes in annual precipitation, conducive to vegetation and fuel reduction or increment, thus there is significant uncertainty in the proximity of desert regions (Fig. 2c).

Clearer conclusions can be drawn from the FPY and PFSL calculation (Figs. 3 and 4). The number of months meeting the set of conditions in Table 1 yields the estimated PFSL (Fig. 3b), and the number of years with at least 1-month meeting the thresholds, the FPY. In the boreal regions, we obtain a general lengthening of the PFS. The PFS of these areas is conditioned by temperature, so the amplified warming of Artic zones⁴⁰ is expected to make the FS longer. Notwithstanding, in certain parts of Eastern Asia, the intense warming is counterbalanced by an increase of the precipitation in certain warm months (see Supplementary Figs. 21 and 22), leading to a slight shortening of our estimated PFS. There is evidence, however, that temperature increases may lead to drier fuels in the future despite the precipitation increase, thus augmenting fire risk, as some investigations have shown for Canada⁴¹. Our results agree in general with several other studies that have previously pointed towards an increase of the FSL in boreal areas^1,17,42, even when some suggest a more pronounced lengthening in more northerly latitudes^1,17. In terms of the frequency of years with fire-prone conditions, the conclusions are even clearer. A general increase of the FPY is observed, especially for northerly latitudes, where the differences reach values of more than +4 years per decade (Fig. 4a). This possible increase in fire activity in boreal areas may result in significant peatland combustion and a release of the large quantities of soil carbon that they store into the atmosphere⁴³. These greenhouse gas emissions may create a positive feedback loop, leading to a further increase in temperature, which in turn will enhance boreal wildfire incidence and more peatland burning.

The Te-dhs fire class, corresponding to temperate climates, would also experience a general lengthening of the PFS (Fig. 3). A future precipitation decline may be especially significant in Southern Europe (Supplementary Fig. 21), associated with an increased anticyclonic circulation yielding more stable conditions⁴⁴, while the temperature rise would be quite homogeneous among all Te-dhs fire-climate class areas. The FS drought intensification around the Mediterranean, together with the general warming (Supplementary Fig. 21), would lead to a lengthening of the PFS of around 2 months (Fig. 3a), but summer months could also experience this precipitation decline (Supplementary Fig. 22), meaning that the FS would be more severe. The Western US, which has already experienced over the last decades the lengthening of the FS⁴⁵ and the increase of large fires⁴⁶ and extreme wildfire weather^47,48 due to climate change, may also experience an FS lengthening by the end of the 21st century. Some authors^18,48,49,50 have studied projected fire future changes from other points of view (occurrence of very large fires, wildfire potential, etc.), finding also a general increase of fire severity by the end of the century in some of these Te-dhs fire regions. The interannual recurrence of fire-prone conditions will significantly increase in countries like France, Italy or Eastern China (Fig. 4a).

The PFSL of the Tropical Tr-ds fire-climate class presents slight differences between present and future values (Fig. 3). Some areas of the Northern African savanna may experience a shortening of the PFS, while Southern Africa shows a lengthening. A dipole pattern of wetting in tropical Eastern Africa and drying in Southern Africa⁵¹ could be the reason for these future changes. There is a contrasting influence of ENSO in present African fire patterns⁵², which suggests that the future pattern of precipitation variations in Central Africa may be associated with ENSO future changes under climate change conditions⁵³. Although the quantification of ENSO changes in a warmer climate is still an issue that continues to be investigated, an expansion and strengthening of ENSO teleconnections is confirmed by some authors^53,54,55. The general increase in precipitation along all seasons in western equatorial Africa would lead to a significant decrease in the recurrence of interannual fire-prone conditions (Fig. 4a).

Our results show that fire-prone areas in Temperate and especially Boreal climates are projected to undergo the most significant expansion and lengthening of the potential FS at the end of the XXI century driven by rising temperatures. In the Tropics, little change is expected in these respects. Notwithstanding, global warming is likely to make fire risk more severe mostly everywhere, and in particular in some regions such as Mediterranean Europe and the Eastern Amazon, where an important decrease in precipitation is also predicted during the PFS. More favourable fire conditions will potentially increment fire activity and burned areas in many places. In others, especially in the Tropics, increasing suppression efforts and a cease to agricultural and pastoral practices like vegetation clearing by fire, replaced by more intensive farming, could counteract the impact of a warmer climate. A reduction of these human-caused fires in the Tropics could bring global burned area down², despite rising trends elsewhere, given the vast contribution of Tropical fires to the burned areas at the global scale (Fig. 1).

Source: Ecology - nature.com