Higher wildfire incidence, severity and population exposure in protected areas within fire-prone Temperate and Mediterranean biomes

doi:10.21203/rs.3.rs-5030414/v1

Download PDF

Article

Higher wildfire incidence, severity and population exposure in protected areas within fire-prone Temperate and Mediterranean biomes

https://doi.org/10.21203/rs.3.rs-5030414/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The European Union has recently passed the Nature Restoration Law (NRL) which, among others, seeks to increase the cover of forest reserves protected for biodiversity and, globally, the Kunming-Montreal Global Biodiversity Framework similarly seeks to expand protected areas. Here we test whether a trade-off exists between protected areas expansion and fire activity, leading to a higher exposure to fire for the population in protected areas, because they often harbour more biomass and occur in remote areas. We analysed forest fires affecting 14,892,174 ha, and intersecting 10,999 protected areas, across fire-prone European Temperate and Mediterranean forest biomes, and in similar ecosystems within California, Chile and Australia. Protected areas were being disproportionally affected by fire within most Temperate biomes, and fire severity was 20% higher within protected areas also in Mediterranean biomes. Population in the periphery of forest areas was up to 16 times more likely to be exposed to large wildfires when their environment was within, or near, protected areas. Enhanced fire activity in protected areas was driven by a combination of fuel loads, accessibility and abiotic factors. Wildfire prevention and mitigation must be central goals in the development of the NRL and other conservation/restoration programs to diminish population exposure and fire severity.

Earth and environmental sciences/Natural hazards

Earth and environmental sciences/Ecology/Forestry

A global pyrocrisis characterized by increases in fire severity, and in the frequency of energetically extreme fire events, is ongoing despite a decline in burned area^1,2. Concomitantly, large-scale land degradation and biodiversity loss are prompting a series of continental legislations and global agreements seeking to expand protected areas, such as national parks and reserves, and to restore degraded ecosystems. But, could these initiatives seeking to preserve, or restore, nature exacerbate the fire problem?

The European Union has recently passed the Nature Restoration Law. Among other goals, this law seeks to increase the cover of protected areas by up to 30% of the land surface. The Restoration Law also specifies that, at the very least, a third of these protected areas should be under strict protection, where no human intervention is allowed. This continental regulation followed after 2022’s Kunming-Montreal Global Biodiversity Framework, a global agreement that seeks to raise the global coverage of protected areas from the current values of 16%, up to 30% by 2030 (the “30×30” program).

Biophysically, fire activity is fundamentally constrained by the production and desiccation of fuels³. The former could increase by the change in status from unprotected to protected, potentially magnifying fire activity. This is because management and human-related disturbances often cease, at least to some degree, after protection. As a result, protected areas have been documented to harbour higher aboveground biomass⁴ which, under drought, may increase the available fuel load and its spatial continuity, therefore increasing fuel hazard. Previous studies have indeed linked an enhanced probability of large fire events with a passive approach to managing forest reserves⁵. However, potential feedbacks between protection status, fire activity and human exposure to fire remain unexplored.

Addressing human exposure to fire is crucial for the success of nature protection and restoration as wildfires can no longer be considered just as an ecological disturbance. They are also becoming an increasing civil protection threat, as the number of fire-induced fatalities in Europe is higher than, for example, the victims of terrorism⁶. Wildfires can have significant impacts on the mental health of exposed citizens⁷, and also on different economic activities⁸, and major fire episodes, such as the unprecedented 2019/20 fire season in southeast Australia⁹, could even threaten national security¹⁰. We thus need to understand whether the expansion of protected areas is affecting population exposure to fire.

If protected areas harbour larger fuel loads than unprotected lands, we can expect that burned area and fire severity will both be higher in protected areas. Fuel load is one of the main drivers of fire intensity¹¹ which, in turn, affects fire severity, or the immediate ecological impacts of fire¹². However, other possible mechanisms could explain the potentially larger impact of fire on protected areas. For instance, protected areas tend to be more remote, and can have lower road densities, which decreases accessibility and complicates firefighting. They could also be more prevalent where other land uses are marginal, such as higher elevation zones, or steeper terrain¹³. Quantitative estimates of the magnitude of the changes in burned area and fire severity in protected vs unprotected areas are currently lacking, and the relative roles of the different possible underlying mechanisms also remain unexplored.

Here, we test whether a surge in fire activity could be expected from the implementation of the Nature Restoration Law. More specifically, we sought to examine if protected areas currently show a positive bias in annual burned area, fire severity and population exposure to wildfire in EU’s most fire-prone forest biomes¹⁴: Mediterranean forests, woodlands and shrublands (MED), and temperate broadleaf and mixed forests (TBLM). We focused on Southwestern Europe (EU; comprising Portugal, Spain and Southern France), as this is a particularly fire-prone area. Additionally, because global efforts such as the 30 × 30 program similarly seek to increase the cover of protected areas, we also examined changes in burned area, fire severity and population exposure between protected and unprotected areas in the MED and TBLM biomes from Southeastern Australia (AUS; comprising New South Wales, Victoria and Tasmania), Chile (CHI) and California (CAL). TBLM is absent from California, and was replaced by the temperate conifer forests biome (CON). Finally, we also examined the mechanisms underlying these patterns and, in particular, whether potential differences in burned area could be driven by fuel loads, road density, fire weather, slope or elevation.

Disproportionate increases in burned area within Temperate, but not Mediterranean, forest biomes

We observed that protected areas burned disproportionately more within the temperate biomes (Table 1, Fig. 1). In Australia, only 18% of the TBLM biome is protected, but 29% of the total burned area in TBLM forests occurred in protected areas. Similarly, protected areas in California’s temperate coniferous forest biome (CON) burn disproportionally: 25% of the total burned area in this biome occurs in protected areas that make up only 17% of the total extent of the CON biome. In Southwest Europe, the proportion of protected areas increased from 22% in 2001 to 38% by 2021. During this time, the proportion of burned area within protected areas increased on average from 13 to 55%. In other words, the extent of protected areas increased significantly over the last 20 years, but the proportion of burned area within protected areas has increased to a significantly larger extent (Fig. 1, Table 1. Chile is the only exception to this trend, as the area under protection increased from 18 to 23% between 2001–2021, while the proportion of burned area within protected areas declined (Fig. 1).

Table 1

**Summary of differences in burned area, fire weather, fire severity and population exposure as function of protection**. The percentage of protected areas within forest areas (%PA in Fig. 1), and the percentage of forest burned area that occurred within protected areas (%BA in PA). β₁ indicates the slope of the regression between %PA (β_{1_PA}) and year (2001–2021), or between %BA in PA (β_{1_BAinPA}) and time. Variation in median fire severity was assessed through the difference in the Normalized Burn Ratio (dNBR) within protected areas (dNBR_{_PA}) and outside protected areas (dNBR_{_NoPA}). We also examined differences in the mean percent of population living in neighbourhoods affected by fire in protected (% Pop_{_PA}) or unprotected (% Pop_{_NoPA}) areas. Error intervals indicate the 95% CI. Significant differences between %PA and %BA in PA (calculated from student’s t), between β_{1_PA} and β_{1_BAinPA} (calculated from slope tests), between dNBR_PA and dNBR_NoPA and between %Pop_PA and %Pop_NoPA (calculated from Wilcoxon signed-rank test) are in bold. Biome names are TBLM for temperate broadleaf mixed biomes, CON for temperate conifer forests and MED for Mediterranean forests, woodlands and shrubs, respectively.
	% PA	% BA in PA	β_{1_PA}	β_{1_BAinPA}	dNBR_{_NoPA}	dNBR_{_PA}	% Pop_{_PA}	% Pop_{_NoPA}
TBLM AUS	17.5 (± 2.0)	28.9 (± 7.1)	0.70 (± 0.04)	0.60 (± 1.17)	0.24 (± 0.02)	0.29 (± 0.02)	0.33 (± 0.29)	0.02 (± 0.02)
TBLM EU	30.3 (± 2.3)	34.3 (± 7.5)	0.80 (± 0.07)	2.10 (± 0.77)	0.29 (± 0.04)	0.36 (± 0.03)	0.12 (± 0.09)	0.03 (± 0.03)
CON CAL	16.6 (± 0.3)	24.5 (± 5.7)	0.11 (± 0.02)	0.05 (± 0.96)	0.28 (± 0.06)	0.31 (± 0.05)	0.22 (± 0.26)	0.06 (± 0.05)
TBLM CHI	21.0 (± 0.8)	11.7 (± 6.0)	0.24 (± 0.07)	-1.04 (± 0.90)	0.35 (± 0.14)	0.36 (± 0.14)	0.17 (± 0.22)	0.11 (± 0.09)
MED AUS	18.8 (± 1.3)	51.6 (± 11.0)	0.42 (± 0.08)	-0.94 (± 1.82)	0.21 (± 0.03)	0.24 (± 0.03)	0.04 (± 0.04)	0.00 (± 0.00)
MED EU	36.5 (± 3.0)	31.7 (± 5.9)	1.05 (± 0.12)	0.96 (± 0.90)	0.32 (± 0.02)	0.39 (± 0.02)	0.05 (± 0.03)	0.01 (± 0.01)
MED CAL	14.5 (± 0.4)	18.2 (± 6.1)	0.14 (± 0.02)	-0.26 (± 1.03)	0.33 (± 0.03)	0.39 (± 0.03)	0.25 (± 0.16)	0.02 (± 0.01)
MED CHI	1.2 (± 0.1)	0.62 (± 0.5)	0.03 (± 0.01)	0.07 (± 0.07)	0.19 (± 0.32)	0.32 (± 0.13)	0.15 (± 0.27)	0.04 (± 0.02)

In the Mediterranean biome of Australia, protected areas occupy 19% of the land but concentrated 52% of the burned area (Table 1, Fig. 1), indicating again that protected areas burn disproportionately. In the Mediterranean biomes of Europe and California, protected areas were affected by fire proportionally to the extent of land they occupy. That is, the increase in protected areas surface over time in Mediterranean Europe (25–45%) was not significantly different from the proportion of burned area occurring within protected areas (22–41%). Similarly, the proportion of protected areas in Mediterranean California (15%) was not significantly different from the proportion of burned area within protected areas (18%). Protected areas were marginal (1.2% of the land) in Mediterranean Chile, and burned area in protected areas was smaller (0.6%) indicating that protected areas were less affected by wildfire than unprotected areas.

Fire severity was often higher within protected areas

We assessed fire severity from the pre- to post-fire change in the remotely sensed Normalized Burn Ratio (dNBR), a commonly used spectral index¹⁵. On average, median fire severity was 21% higher inside protected areas within the temperate biome (Fig. 2, Table 1). The median annual dNBR ranged from 0.24 outside protected areas to 0.29 inside protected areas in AUS, and from 0.29 to 0.36 in EU. In CAL, dNBR was only marginally higher in unprotected areas (0.28) relative to protected areas (0.31) and, in Chile, the difference was also not significant (0.35–0.36).

Regarding the Mediterranean biome, median fire severity was 19% higher inside than outside protected areas, and differences were significant in AUS (0.21 in unprotected and 0.24 protected areas), CAL (0.33 in unprotected and 0.39 in unprotected areas) and in the EU (0.32 in unprotected and 0.39 in unprotected areas). The only exception was Chile, where the median dNBR was not significantly different inside and outside protected areas.

Population exposure to wildfire was consistently higher in protected areas

We also observed that, considering the rural and Indigenous communities living in the studied regions, a larger proportion of the population had been exposed to fires burning within protected areas, relative to those burning outside protected areas (Table 1). The overall proportion of people annually exposed to fire was always smaller than 0.5%, as expected given the low population density in forest regions, and that fires often affect only a small percentage of the land. However, averaged across years, it was up to 16 times more likely that population had been affected by a fire within protected areas, relative to those burning outside protected areas. Within the temperate forest biomes, there were significant differences in AUS and EU, where the percentage of the population affected by fire within protected areas ranged from 0.12–0.33%, depending on continent. Outside protected areas, however, these percentages dropped to 0.02–0.03%.

Within the Mediterranean biome, there were also significant differences in population exposure to fires across protection status within AUS, EU and CAL. Here, the proportion of the population affected by fire within protected areas was also larger (0.04–0.25%) than for fires outside protected areas (0.0001-0.02%) (Table 1). It is worth noting that, in AUS and EU, the ratio between the proportion of the population exposed to fire inside protected areas, relative to those outside protected areas, has significantly increased over the last 20 years (Fig. 3). In other words, while it has been more likely for a citizen to be exposed to fires burning within a protected area at any given time during the study duration, this difference has been widening over time. In Australia, it was 7 times more likely to be exposed to a fire inside than outside a protected area in 2001, while this ratio increased to 44 by 2020 and, in Europe, this ratio increased from 1.4 in 2004 to 9 by 2020 (Fig. 3). It is worth noting that these results are independent from changes in population, as they are expressed on a relative basis.

Differences in burned area driven by fuel load, accessibility and abiotic factors

Differences in burned area across land-use types were driven, at least partly, by changes in fuel loads in both biomes. We used net primary productivity (NPP) in the year prior to the fire as a proxy for biomass accumulation¹⁶, and we observed a positive association with the difference in burned area across protection status in the Mediterranean biome (Fig. 4). In other words, increasingly higher differences in burned area in protected areas, relative to unprotected areas, were significantly and positively associated with significantly higher NPP in protected areas, relative to unprotected areas. Additionally, changes in altitude and vapour pressure deficit (VPD), and the three-ways interaction between VPD, NPP and road density, all affected the difference in burned area between protected and unprotected areas. In the Temperate biome, we similarly observed that the three-ways interaction between VPD, NPP and accessibility showed the largest, in absolute terms, effect size over the change in burned area. Additionally, interactions between elevation and accessibility, and between NPP, accessibility and steepness, also affected significantly the difference in burned area between protected and unprotected areas (Fig. 4).

We observed a trade-off between nature protection (for biodiversity conservation) and fire risk to local communities. The proportion of rural and Indigenous communities exposed to fire within protected areas was up to 16 times larger, depending on biome and region, relative to those living in unprotected lands. This trend has increased significantly in Australian and European forests during 2001–2021, and it has remained stable elsewhere. It has thus become increasingly more dangerous to live within a protected area, or its periphery, over the last two decades in Australia and southwestern European Union.

Plans for expanding protected areas under EU’s Nature Restoration Law, and by the global 30×30 program, need to address this concern. Protected areas must prioritise the safety of the population living within, or near, those environments. To this end, and in order to develop appropriate policies, we need to understand the mechanisms underlying this increase in fire exposure.

Differences in burned area across land of different protective status were driven by higher fuel loads, along with changes in accessibility (road density) and abiotic factors. Abiotic factors cannot be managed, but hazard can be mitigated through fuel management. EU’s Nature Restoration Law seeks to increase the proportion of land under strict protection. Previous research has shown that aboveground forest biomass is larger in protected areas⁴ and it is to be expected that future increases will be the largest under strict protection, all else being equal, if human intervention for managing fuels is not allowed. At the very least, potential fire activity should be an integral aspect to consider in the establishment and management of protected areas and, particularly, in areas under strict protection. Otherwise, a completely passive management could lead to unintended increases in fire activity⁵, which would be a perverse outcome of the Restoration Law. Appropriate road access may also be needed, if not everywhere, at least in critical areas to ensure firefighters can defend communities.

The only relative decline in burned area within protected areas occurred in Chile. In Mediterranean Chile, protected areas occupy only 1.2% of the land, so it is too early to assess the impacts of protection over fire activity. In temperate Chile, wildfire is often associated with land use conflicts over plantations¹⁷, which could explain the lower impact over protected areas. Additionally, protected areas in temperate Chile are distributed towards the southern end, which is cooler and wetter (Supplementary Fig. 1). However, we argue that fire management should also be applied in those protected areas, in order to avoid the problems that we document here for AUS, CAL and EU under a warming climate. This year’s record-setting Valparaíso fire, for instance, which caused 133 fatalities, started in a protected area (Lago Peñuelas National Reserve) and burned substantially through native and protected forest ¹⁸.

In addition to a bias in burned area and population exposure, fire severity was often higher in protected areas than in non-protected areas, which could jeopardize conservation efforts within protected areas. Low-intensity fire regimes generate pyrodiversity^19,20, but anthropogenic regimes with a growing presence of extreme fires tend to homogenize landscapes and, consequently, reduce biodiversity²¹. Increases in fire severity have been linked with decreases in plant abundance, diversity and fitness ²². If protected areas favour high severity fires, this might counteract any protection efforts, especially for fire-sensitive species. Fire management plans should thus be an integral component also in protected areas, in order to maximize the benefits of fire, and minimize its negative effects. If not, the global extreme wildfire problem, which is projected to increase under ongoing climate change²³, may be further exacerbated by the expansion of protected areas.

The higher biomass in protected areas could result from a restriction of human disturbances after the area was protected. However, it is also possible that areas that are now protected already had higher biomass levels before protection was established. Regardless of the reason, removal of hazardous biomass is our primary tool to diminish population exposure to fire. We are not advocating for fire prevention treatments to be implemented everywhere within protected areas, but call for adequate risk mitigation measures to protect local communities as a primary goal. Protection for biodiversity does not necessarily protect against wildfire.

Several options exist for reducing fuels²⁴. Prescribed and cultural burning are often considered as the most effective strategy, as they remove both live and dead fuels, but they need to be applied over large scales to be effective^25,26. It is often found that prescribed fire is even more effective when accompanied by thinning that follows pyro-silvicultural principles²⁷. Additionally, herbivores, either domestic or wild, may be effective to control live fuels, especially in grasslands and shrublands. Green fuelbreaks are also important at the wildland-urban interface²⁸. The variety of management techniques that can be applied to reduce fire hazard is wide, and performing a full review would be beyond scope. It is worth noting here that most of these strategies seek to mimic nature, and can be implemented without compromising the “naturalness” that is expected from protected areas. At any rate, management actions should be tailored to the specific conditions of each site.

It is worth mentioning that EU’s Restoration Law does mention that “Member States should take into account the risks of forest fire”. However, the Commission’s Guidelines for land-based wildfire prevention²⁹ recommends, among others, developing multi-species and multi-layered forest structures “since their more complex structure can slow fire spread” relative to simple, mono-specific forests. Our study was not designed to test this assumption as such. However, we can expect that protected areas currently show a higher proportion of complex forest structures than unprotected areas. Thus, our finding that protected areas often burned under higher severity, and that the proportion of burned area was often equal, or higher, in protected areas, does not support the assumed role of complex vegetation structures in burned area mitigation.

A plethora of well-intended policies are currently being implemented in Europe and in global initiatives such as the 30×30 plan. At best, restoring native forests and expanding protected areas will likely bring a proportional increase in burned area, while disproportionate increases are already observed in some regions from Australia, California and Europe, with important implications on fire severity and population exposure. Well-intended policies can only be effective when they are based on broad, science-based assessments of the pros and cons of each measure, and on solid guidelines. Nature protection should not occur at the expense of increase population exposure to wildfire.

Site selection

Our study region comprises three forest biomes on four continents (Australia, Europe, North America, South America). Within each continental biome climatic conditions are similar, but the percentage of protected areas varies markedly. Southwestern EU has the highest coverage of protected areas in semi-natural vegetation, near or over the 30% depending on the country, while only 1.2% of the Mediterranean forest biome in CHI is protected (Supplementary Fig. 1). Cover of protected areas is intermediate (14.5–18.8%) in AUS and CAL. National Parks in California were excluded (“w/o Nat. Parks”) because their wildland fire-use policy, whereby natural fires are managed for resource benefits³⁰, could increase burned area in protected areas. We sought regions with Mediterranean and Temperate biomes and South Africa, which only has Mediterranean ecosystems, was not included.

To ensure that the actual vegetation types within the forest biomes were comparable, we restricted our analyses to semi-natural vegetation, which includes forests, grasslands, shrublands and sparse vegetation, as classified in the European Space Agency (ESA) Climate Change Initiative (CCI) medium-resolution land cover (MRLC) dataset v 2.0.71³¹. For simplicity, we refer to this vegetation as “forests”. We excluded land cover types, like crops, that seldom burn. With this mask, we avoided artefacts related to a higher presence of non-flammable crops outside protected areas, which could artificially increase the impact of fires within protected areas. This mask was applied considering the conditions of the year before a given fire. That is, the analysis of a fire occurring in 2020 would be based on the value of the mask released at the end of 2019.

Data sources

We used GlobFire v.3, a global burned-area dataset at 500m resolution (2001–2021)³², which is based on MODIS Collection 6.1 (C6.1), MCD64A1 burned area product. To understand the effects of fire weather, we examined patterns in the aggregated vapor pressure deficit (VPD). At seasonal scales, VPD is an indicator for live and dead fuel moisture and a key driver of fire weather that is becoming widely used over several studies³³. VPD was calculated using the ERA5-Land reanalysis ³⁴, available in Google Earth Engine, at 15.00h local solar time on the day of fire detection, resampled by bilinear interpolation matching 500m MODIS resolution. 15.00h was chosen following previous publications³⁵, as this is often the hottest and driest part of the day.

Data on protected areas coverage and year of establishment were obtained from the World Database of Protected Areas (WDPA), a joint project between the United Nations Environment Programme (UNEP) and the International Union for Conservation of Nature (IUCN) ⁴¹. This database is considered to be the most comprehensive global database of protected areas and is updated monthly. In this study, we used the November 2023 version of the WDPA database, and we included 10,999 protected areas, covering a total of 37,700,585 ha. WDPA also includes the IUCN management category (where human intervention increases from I-VI). However, this data was not available for all countries and therefore could not be included in our analyses. The database relies on national contact points, and we urge them to make this information open access, as that will be important for further analyses of fire regimes within protected areas. Following previous studies⁴², we generally adopted the WDPA best practice guidelines, but we included legally-designated sites only. We followed Butchart et al⁴³ when the status year was not available. That is, we assigned missing establishment dates by randomly selecting a year (with replacement) from all protected areas within the same country with a known date of establishment. For countries with fewer than five protected areas with known year of establishment, a year was randomly selected from all terrestrial protected areas with a known date of establishment. The random assignment was repeated 1,000 times, to identify the median and year of establishment, which we assigned to each protected area without an establishment date.

We assessed fire severity after calculating dNBR using the same methods as MOSEV³⁶, a global product of dNBR:

\(\:NBR=\frac{\rho\:2-\rho\:7}{\rho\:2+\rho\:7}\) (Eq. 1)

\(\:\text{d}\text{N}\text{B}\text{R}\:=\:\text{p}\text{r}\text{e}\text{N}\text{B}\text{R}\:-\:\text{p}\text{o}\text{s}\text{t}\text{N}\text{B}\text{R}\) (Eq. 2)

where \(\:\rho\:2\) and \(\:\rho\:7\) are the land surface reflectance values of bands 2 (841–876 nm) and 7 (2105–2155 nm) from MODIS Terra MOD09A1 and Aqua MYD09A1 collection 6.1 data products available in GEE. We kept only the highest quality pixels (e.g., removing cloud effects and water bodies) for each individual band using the respective MOD09A1 and MYD09A1 bitmask quality assessment bands. As in MOSEV, “Terra NBR gaps (masked areas) were re-filled with the synchronous Aqua NBR values when available” to increase data availability. We selected the immediate preNBR and postNBR based on the initial and final fire dates provided by the GlobFire database, respectively, and considering the temporal resolution of 8 days from the Terra-Aqua NBR composites. When NBR values were not available at pixel level, we filled them with the previous or the next image until a limit of 40 days (that is, a maximum of 5 images). This limit captures the closest pre- and post- fire conditions before the onset of significant changes in the vegetation (e. g., resprouting). dNBR median values were then calculated across conservation status (protected and unprotected areas). All calculations were made in Google Earth Engine³⁹.

Population data was obtained by combining the unconstrained top-down population dataset from WorldPop ³⁷, with the official population values from the UN ³⁸. Analyses were performed within Google Earth Engine and the R environment v. 4.4.0 ⁴⁰.

The analysis of the drivers underlying the difference in burned area between protected and unprotected lands required data on fuel loads, road density, VPD, slope and elevation. Fuel load data was estimated using Net Primary Productivity (NPP) the year before the fire as a proxy for fuel accumulation, following previous approaches¹⁶. Yearly NPP was extracted from MODIS MOD17A3HGF Collection 6.1 at 500-m resolution available in GEE. Road density (km km^− 2) was obtained from the Global Roads Inventory Project (GRIP) dataset⁴⁴, and slope (º) and elevation (m) from the Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010)⁴⁵ .

Analyses of burned area

Our analyses are based on a total of 76,861 fires, where 12,424 fires (16.2%) affected protected areas. Total burned area was 14,892,174 ha, with 4,483,045 ha (30.1%) occurring in protected areas, and 10,409,129 ha (69.9%) in unprotected areas (Supplementary Fig. 1). We assessed the effects of protection on burned area based on two separate analyses. First, we examined the percentage of land covered by protected areas within forest areas (%PA in Fig. 1), and the percentage of burned area that occurred within protected areas (%BAinPA in Fig. 1) during the study period (2001–2021). If protected areas burn preferentially, then we should observe that the percentage of burned area within protected areas is larger than the percentage of land that is protected in each of the biomes (%BAinPA > %PA). For example, if protected areas occupy 10% of the forest area within a biome, but 50% of the burned area occurs in protected areas (%BAinPA > %PA), then fires would be disproportionately affecting protected areas.

If there is no effect of protection on burned area, we expected that %PA would not be significantly different from %BAinPA (%BAinPA = %PA). Finally, if protected areas are less affected by fire than the rest of the landscape, then the percentage of the area burned within protected areas should be smaller than the percent of land occupied by protected areas (%BAinPA < %PA). Statistical analyses on differences between annual values of %BAinPA and %PA were assessed with student’s t after testing for normality assumptions.

In the second analysis, we assessed temporal changes in the slope of %PA, and %BAinPA. This analysis was necessary because, in some regions, %PA was not stable, but it increased continuously through time. Thus, we sought to understand whether or not %BAinPA had increased at the same pace as %PA using slope tests ⁴⁶. If the slope of the temporal change in %BAinPA was larger than that in %PA, then we would be observing a disproportionate effect of fire in protected areas. Conversely, if increases in %BAinPA are more modest than those in %PA, then protected areas would be less affected by fire than non-protected areas.

Analyses of fire severity and population exposure

In order to test whether fires burning within protected areas were more severe, we restricted our analyses to large fires (> 500ha, which represent the bulk of the burnt area) that had a minimum of 20% of their burned area within a protected area (to ensure representativeness of protected areas in the sample). We subsequently intersected protected areas data from WDPA with burned area data from GlobFire, and extracted dNBR median pixels for each large fire, separately within and outside the protected area. Our analysis was thus a paired approach, comparing intra-fire variation in dNBR depending on protection status for each pixel. Statistical differences in dNBR within and outside protected areas were then analysed with non-parametric Wilcoxon signed-rank tests, as the data were not normally distributed.

Population exposure was defined as the presence of population in pixels that burned. This does not include populations exposed to smoke or those in proximity to the fires who might be indirectly impacted, as that would require making additional assumptions. Quantitatively, any pixel from the population dataset, along with the number of people residing in that pixel, were considered exposed if the centroid of the pixel was within the fire perimeter. We quantified all the population affected by large fires (> 500ha), and we quantified intra-fire variation in population exposure depending on whether population affected was in protected or unprotected areas. We calculated the percentage of the population exposed to fire in protected areas, relative to the total population in protected areas, and the percentage of the population exposed to fire in unprotected areas, relative to the total population in unprotected areas. Total burned area is, overall, higher in unprotected areas because they occupy a larger extent of the land. This approach thus tends to underestimate population exposure in protected areas, as total burned area in protected areas was only 30.1% of total burned area. This approach was chosen because it measures actual population exposure, but it should be noted that actual effects of protection on population exposure would likely be higher than those reported here, if expressed on a burned area basis. Statistical differences were assessed through non-parametric Wilcoxon signed-rank tests, as the data were not normally distributed.

We also wanted to test for differences in the proportion of the population affected by fire in protected and unprotected forests, and whether this pattern had changed over time. We examined the ratio between the proportion of the population that lives in protected areas and has been affected by a fire, relative to the proportion of the population that lives outside protected areas and was affected by a wildfire, within each year. Population affected was, in some biomes/years inferior to 50 people and, in that case, data for that particular biomes/year was not considered in subsequent quantitative assessments (only 1 outlier was discarded). A linear regression analysis was carried out to assess the trend over time.

Analyses of the drivers underlying increases in burned area

In order to understand the factors underlying the annual changes burned area at each biome/region, we fitted linear mixed models where DBA (the difference in burned year in protected areas minus that in unprotected areas) was the dependent variable. The fixed factors were DNPP the year prior to the fire, D VPD on the day of fire detection, D elevation, D slope and D road density, and their interactions within a full factorial model. Similar to dNBR analyses, we focused on large fires that affected protected areas substantially (20–80% of the total burned area in protected areas). Significant differences across coefficients were assessed through Wald tests. The values of the independent variables were re-scaled prior to analyses so that coefficient estimates were comparable and provided effective information on the effect size ⁴⁷. These analyses were performed in R (v.4.4.0) using the lme4 ⁴⁷ and car ⁴⁸ libraries. Analyses were performed after assessing homocedasticity and normality in the residual plots, and the dependent variable was log-transformed to ensure normality assumptions were met if required.

Acknowledgements

This work was supported by the EU H2020 program (grant agreements 101003890 and 101037419) and the Spanish MICINN (PID2022-138158OB-I00). PMF was supported by the Portuguese Foundation for Science and Technology through project UIDB/04033/2020 (DOI: 10.54499/UIDB/04033/2020). R.B-R was funded by the UNED-SANTANDER grant program. We appreciate feedback and earlier discussions with F Castañares, J Saavedra, M Castellnou and R Domènech.

Cunningham, C. X., Williamson, G. J. & Bowman, D. M. J. S. Increasing frequency and intensity of the most extreme wildfires on Earth. Nat Ecol Evol (2024) doi:10.1038/s41559-024-02452-2.
Fernández-García, V. & Alonso-González, E. Global Patterns and Dynamics of Burned Area and Burn Severity. Remote Sensing 15, 3401 (2023).
Boer, M. M., Resco De Dios, V., Stefaniak, E. Z. & Bradstock, R. A. A hydroclimatic model for the distribution of fire on Earth. Environ Res Comm 3, 035001 (2021).
Duncanson, L. et al. The effectiveness of global protected areas for climate change mitigation. Nat Commun 14, 2908 (2023).
Agee, J. K. The fallacy of passive management: managing for firesafe forest reserves. Conservation in Practice 3, 18–26 (2002).
Resco de Dios, V. & Nolan, R. H. Some Challenges for Forest Fire Risk Predictions in the 21st Century. Forests 12, 469 (2021).
Laugharne, J., Van de Watt, G. & Janca, A. After the fire: the mental health consequences of fire disasters. Current Opinion in Psychiatry 24, (2011).
Meier, S., Elliott, R. J. R. & Strobl, E. The regional economic impact of wildfires: Evidence from Southern Europe. Journal of Environmental Economics and Management 118, 102787 (2023).
Boer, M. M., Resco de Dios, V. & Bradstock, R. A. Unprecedented burn area of Australian mega forest fires. Nature Clim Change 10, 171–172 (2020).
McDonald, M. After the fires? Climate change and security in Australia. Australian Journal of Political Science 56, 1–18 (2021).
Byram, G. M. Combustion of forest fuels. in Forest Fire: Control and Use (ed. K. P. Davis) 61–89 (McGraw-Hill, New York, 1959).
Keeley, J. E. Fire intensity, fire severity and burn severity: a brief review and suggested usage. Int J Wildland Fire 18, 116–126 (2009).
Baldi, G., Texeira, M., Martin, O. A., Grau, H. R. & Jobbágy, E. G. Opportunities drive the global distribution of protected areas. PeerJ 5, e2989 (2017).
Dinerstein, E. et al. An Ecoregion-Based Approach to Protecting Half the Terrestrial Realm. Bioscience 67, 534–545 (2017).
Miller, J. D. & Thode, A. E. Quantifying burn severity in a heterogeneous landscape with a relative version of the delta Normalized Burn Ratio (dNBR). Remote Sensing of Environment 109, 66–80 (2007).
Galizia, L. F., Curt, T., Barbero, R. & Rodrigues, M. Understanding fire regimes in Europe. Int. J. Wildland Fire (2021) doi:10.1071/WF21081.
Gómez-González, S., Ojeda, F. & Fernandes, P. M. Portugal and Chile: Longing for sustainable forestry while rising from the ashes. Environmental Science & Policy 81, 104–107 (2018).
González, M. E., Syphard, A. D., Fischer, A. P., Muñoz, A. A. & Miranda, A. Chile’s Valparaíso hills on fire. Science 383, 1424–1424 (2024).
Martin, R. E. & Sapsis, D. B. Fires as agents of biodiversity: pyrodiversity promotes biodiversity. Proceedings of the Symposium on Biodiversity of Northwestern California (1992).
He, T., Lamont, B. B. & Pausas, J. G. Fire as a key driver of Earth’s biodiversity. Biological reviews of the Cambridge Philosophical Society 94, 1983–2010 (2019).
Karavani, A. et al. Fire-induced deforestation in drought-prone Mediterranean forests: drivers and unknowns from leaves to communities. Ecological Monographs 88, 141–169 (2018).
Grau-Andrés, R., Moreira, B. & Pausas, J. G. Global plant responses to intensified fire regimes. Global Ecol Biogeogr 33, e13858 (2024).
Clarke, H. et al. Forest fire threatens global carbon sinks and population centres under rising atmospheric water demand. Nat Commun 13, 7161 (2022).
Agee, J. K. & Skinner, C. N. Basic principles of forest fuel reduction treatments. Forest Ecology and Management 211, 83–96 (2005).
Boer, M. M., Sadler, R. J., Wittkuhn, R. S., McCaw, L. & Grierson, P. Long-term impacts of prescribed burning on regional extent and incidence of wildfires—evidence from 50 years of active fire management in SW Australian forests. For Ecol Manag 259, 132–142 (2009).
Davim, D. A., Rossa, C. G., Pereira, J. M. C. & Fernandes, P. M. Evaluating the effect of prescribed burning on the reduction of wildfire extent in Portugal. Forest Ecology and Management 519, 120302 (2022).
Davis, K. T. et al. Tamm review: A meta-analysis of thinning, prescribed fire, and wildfire effects on subsequent wildfire severity in conifer dominated forests of the Western US. Forest Ecology and Management 561, 121885 (2024).
Bowman, D. M. J. S. & Sharples, J. J. Taming the flame, from local to global extreme wildfires. Science 381, 616–619 (2023).
European Commission, Directorate-General for Environment, Nuijten, D., Onida, M. & Lelouvier, R. Land-Based Wildfire Prevention – Principles and Experiences on Managing Landscapes, Forests and Woodlands for Safety and Resilience in Europe. (Publications Office of the European Union, 2021). doi:10.2779/695867.
Stephens, S. L. et al. Fire, water, and biodiversity in the Sierra Nevada: a possible triple win. Environ. Res. Commun. 3, 081004 (2021).
Defourny, P. et al. Observed annual global land-use change from 1992 to 2020 three times more dynamic than reported by inventory-based statistics. in preparation (2023).
Artés, T. et al. A global wildfire dataset for the analysis of fire regimes and fire behaviour. Scientific data 6, 296 (2019).
Clarke, H. Forest fire threatens global carbon sinks and population centres under rising atmospheric water demand. Nature Communications 13, 7161 (2022).
Muñoz-Sabater, J. et al. ERA5-Land: a state-of-the-art global reanalysis dataset for land applications. Earth Syst. Sci. Data 13, 4349–4383 (2021).
Resco de Dios, V. et al. Convergence in critical fuel moisture and fire weather thresholds associated with fire activity in the pyroregions of Mediterranean Europe. Science of The Total Environment 151462 (2021) doi:10.1016/j.scitotenv.2021.151462.
Alonso-González, E. & Fernández-García, V. MOSEV: a global burn severity database from MODIS (2000–2020). Earth Syst. Sci. Data 13, 1925–1938 (2021).
WorldPop (www.worldpop.org - School of Geography and Environmental Science, University of Southampton; Department of Geography and Geosciences, University of Louisville; Departement de Geographie, Universite de Namur) & Center for International Earth Science Information Network (CIESIN), Columbia University. Global High Resolution Population Denominators Project - Funded by The Bill and Melinda Gates Foundation (OPP1134076). https://dx.doi.org/10.5258/SOTON/WP00645 (2018).
United Nations, Department of Economic and Social Affairs, Population Division. World Population Prospects 2022 online edition. https://population.un.org/wpp/ (2022).
Gorelick, N. et al. Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sensing of Environment 202, 18–27 (2017).
R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/. (2024).
UNEP-WCMC & IUCN. Protected Planet: The World Database on Protected Areas (WDPA) and World Database on Other Effective Area-based Conservation Measures (WD-OECM) [Online], November 2023, Cambridge, UK: UNEP-WCMC and IUCN. Available at: www.protectedplanet.net. (2023).
Maxwell, S. L. et al. Area-based conservation in the twenty-first century. Nature 586, 217–227 (2020).
Butchart, S. H. M. et al. Shortfalls and Solutions for Meeting National and Global Conservation Area Targets. CONSERVATION LETTERS 8, 329–337 (2015).
Meijer, J. R., Huijbregts, M. A. J., Schotten, K. C. G. J. & Schipper, A. M. Global patterns of current and future road infrastructure. Environ. Res. Lett. 13, 064006 (2018).
Danielson, J. J. & Gesch, D. B. Global Multi-Resolution Terrain Elevation Data 2010. 34 (2011).
Andrade, J. M. & Estévez-Pérez, M. G. Statistical comparison of the slopes of two regression lines: A tutorial. Analytica Chimica Acta 838, 1–12 (2014).
Bates, D., Maechler, M., Bolker, B. & Walker, S. Fitting Linear Mixed-Effects Models using lme4. Journal of Statistical Software 67, 1–48 (2015).
Fox, J. & Weisberg, S. An R Companion to Applied Regression. (Sage Publications, Thousand Oaks, CA, USA, 2019).

There is NO Competing Interest.

Download PDF

Version 1

posted

You are reading this latest preprint version

Higher wildfire incidence, severity and population exposure in protected areas within fire-prone Temperate and Mediterranean biomes

Status:

Version 1

Abstract

Figures

Introduction

Results

Disproportionate increases in burned area within Temperate, but not Mediterranean, forest biomes

Fire severity was often higher within protected areas

Population exposure to wildfire was consistently higher in protected areas

Differences in burned area driven by fuel load, accessibility and abiotic factors

Discussion

Methods

Site selection

Data sources

Analyses of burned area

Analyses of fire severity and population exposure

Analyses of the drivers underlying increases in burned area

Declarations

References

Additional Declarations

Status:

Version 1