Forest fires: European scale
Fire regimes are commonly characterized by burn frequency and severity within a given area. Severity is often estimated as the proportion of overstory trees killed by fire. In general, as frequency increases, fuels have less time to accumulate, reducing intensity and subsequent tree mortality. Increased frequency and size of large, severe forest fires are expected in Australia, the Mediterranean Basin, Canada, Russia, and the United States (1). The critical issue is whether tree mortality patch sizes (and their temporal and spatial frequency) allow recovery of the same or similar vegetation types. If high-severity patch sizes are too large, microclimates and regeneration mechanisms (e.g., seed abundance and dispersal) can limit tree reestablishment This may result in undesirable ecosystem changes. Rising temperatures, related drought stresses, and increased fuel loads are driving high-severity patches to extraordinary sizes in some areas (2).
In Southern Europe, fire frequency and wildfire extent significantly increased after the1970s compared with previous decades (6) due to fuel accumulation (7), climate change (8) and extreme weather events (9) especially in the Mediterranean basin (10). Future wildfire risk is projected to increase in Southern Europe (11), with an increase in the occurrence of high fire danger days (12) and in fire season length (13). The annual burned area is projected to increase by a factor of 3 to 5 in Southern Europe compared to the present under the A2 climate change scenario by 2100 (14). In Northern Europe, fires are projected to become less frequent due to increased humidity (15).
Wildfire changes the hydrologic response of watersheds, increasing the potential for runoff and erosion relative to unburned conditions (17).
Fingerprint global warming on fire conditions
It is to be expected that wildfires will occur more frequent and will become more severe as a result of global warming. In fact, increases in the frequency and severity of fire weather have been observed across the globe over the past half-century. These increases are not necessarily due to global warming. Other anthropogenic activities, such as land-cover change, population growth, and changes in fire suppression, may play a role as well.
More insight into the attribution of climate change to fire weather conditions is important: it highlights potential timeline for adaptation measures. This calls for specific analysis to distinguish the impact of anthropogenic climate change from other factors, and from natural variability. These analyses can also look into the future and distinguish from climate model simulations when the fingerprint of climate change on fire weather conditions has become so strong that we can actually say that anthropogenic climate change is indeed affecting wildfires. These specific analyses are called attribution studies.
On a global scale, the emergence of anthropogenic climate change in fire weather conditions was studied by using a large number of climate models and simulating four climate characteristics that influence fire risk: daily maximum temperature, daily minimum relative humidity, daily accumulated precipitation, and daily mean wind speed (43). These four variables can be combined into an index called Fire Weather Index (FWI), an indicator of potential fire intensity. This indicator combines the impacts of fuel aridity and fire weather, and has been shown to be a good indicator for burned area across broad regions of the globe. Changes in this indicator can be expressed in different metrics that indicate changes in the frequency of occurrence of extreme fire weather days, in the length of the fire weather season, and in upper values of the risk level. Projected changes in future decades were compared with the period 1861-2005; the latter period describes natural variability of (more-or-less) preindustrial conditions. The future projections were based on a high-end scenario of climate change (the so-called RCP8.5 scenario).
Climate change impacts already observed: The results show that for some of these metrics a climate change signal is already emerging for parts of the globe, including much of southern Europe and the Amazon. Global areas showing this emergence will expand with continued warming over the twenty-first century. These findings suggest substantial increases in fire potential in regions with abundant vegetation (fuel). By 2050, climate change will have increased the frequency of occurrence of extreme fire weather days for 62% of the world’s burnable land. Due to climate change the length of the fire weather season will have increased for 53% of the world’s burnable land, and the upper values of the fire risk level will have increased for 34% of the world’s burnable land. In fact, already in 2019, for 22% of global burnable land surfaces climate change has clearly increased the frequency of occurrence of extreme fire weather days (43).
Mediterranean hot spot: The Mediterranean stands out as a region where wildfire risk increases fast: by 2030, both the frequency of extreme fire weather days, the upper values of risk levels, and the length of the fire weather season will have increased so much that a relationship with climate change can be clearly established. This is consistent with the development of this region as a hot spot of increasing drought risk (44).
Impact Paris Agreement: What if we succeed in stopping global warming at 2°C (the Paris Agreement)? According to this study the local impact of climate change on the frequency of extreme fire weather days and the length of the fire season would still be evident for 30 and 21% of global burnable land, respectively. Most pronounced in the Mediterranean, southern Africa, and portions of the Americas. These percentages would nearly double at 3°C warming.
Changing wildfire regimes in Mediterranean Europe
The impact of 1.5, 2, and 3°C global warming
Many recent studies on the impact of climate change focus on the impact of 1.5, 2, and 3°C global warming. The first two levels of warming refer to the targets of the Paris Agreement. The third level is the global warming the world is currently heading for. The impacts of these three levels were also studied with respect to future wildfires in Mediterranean Europe. This was done by linking climate and summer burned area, and projecting these relationships for different climate scenarios. In these projections, changes in fuel productivity under climate change, that might affect the climate-fire relationships, were also taken into account. The latter refers to the fact that wetter areas produce more vegetation. Thus more fuel is available in these areas in times of droughts, making these wetter areas more fire prone than drier regions in dry periods (39).
Burned area in Mediterranean Europe has not increased in recent decades. Apparently, management actions have so far counterbalanced the climatic trend of increasing drought conditions (18). This will change in future decades. Higher levels of global warming are projected to increase drought conditions that in turn lead to larger burned areas. Limiting global warming to 1.5°C can strongly reduce the increase of burned area (39).
The results show 40-54% increase of burned area over Mediterranean Europe under 1.5°C global warming. This increases to 62-87% for 2°C, and 96-187% for 3°C global warming. These findings substantially align with the results of previous studies that assessed the impact of climate change on burned area in the Mediterranean Basin (40). The study also indicates that a higher level of global warming reduces the productivity of vegetation, and therefore fuel, which limits the sensitivity of fire activity to dry periods. These results do not consider future changes in fire management policies, land-use and land-cover change.
According to these findings, burned area will increase in Mediterranean Europe. In combination with the increase in societal exposure to large wildfires in recent years (41), this calls for a rethinking of current management strategies (42). Climate change effects could overcome fire prevention efforts, implying that more fire management efforts must be planned in the near future. The trend of reduction of burned area in Mediterranean Europe in the past few decades can be explained by an increased effort in fire management and prevention. However, keeping fire management actions at the current level might not be sufficient to balance a future increase in droughts (39).
Human activities versus climate change
Higher temperatures and more droughts not necessarily lead to an increase in the number of wildfires or the area burnt annually. Human activity may alter fire regimes to such an extent that climate change impacts are completely overruled. These human alterations may both increase and decrease fire probability (18). An increase may result from land-use changes or more outdoor activities (including tourism) (19). A decrease may result from more effective fire suppression. Changes in land-use and fire suppression policy seem to have exceeded the strength of climate change effects on changing fire regime in southern France during the period 1976-2009. The complex interaction of climate change and human effects, which may vary regionally and from one season to another, makes regional predictions of future fires highly challenging (18).
Climate effects: fuel-limited versus drought-driven fire regimes
With respect to climate effects, the situation is even more complicated because climate affects wildfires in two opposing ways. In dry ecosystems wet conditions may be so rare that not enough fuel accumulates to start large fires, and fire activity is limited. These areas have a fuel-limited fire regime. In moist ecosystems on the other hand, dry conditions may be so rare that fuel doesn’t get sufficiently dry to sustain fire spread. These areas have a drought-driven fire regime (18). The alternation of wet and dry conditions, however, favours fires by increasing the amount of fine fuel in litter, grass and shrub layers during wet periods, which burn more intensely in subsequent dry periods (20). In Mediterranean ecosystems both fuel-limited and drought-driven fire regimes are present (21).
Human alterations of fire regimes
Human activities directly modify the fire regime In the Mediterranean Basin by setting or suppressing fires and by changing patterns of vegetation in the landscape (22). Moreover, the fuel build-up following agricultural land abandonment as a result of the rural exodus has created an increasing fire hazard in Mediterranean Europe (23). On the other hand, changes in fire suppression policy over the last few decades have probably induced sharp decreases in fires (24). Hence the functional relationships linking fire to climate have been partially modified by human activities (25), decreasing or increasing the fire activity independently of climate change.
The example of southern France
In southern France three pyroclimates can be discriminated (18): (1) the Mediterranean mountains, characterized by a high seasonality in precipitation and fires (dry summers and wet winters), and the highest burned area fraction, fire season length and the strongest increase in fire danger over the last four decades, (2) the Temperate mountains, characterized by wet and cold conditions, and the shortest fire seasons and the lowest fire activities, and (3) the Mediterranean lowlands, with the driest and warmest climates, and strong and dry winds (the Mistral and Tramontane) that favour fire spread in summer. Human presence and activities decrease from the Mediterranean coast to the rural hinterland, with anthropogenic ignitions accounting for 90% of the number of fires and 96% of the total burned area (26).
During the period 1976-2009 the climatic influence on fires in southern France was not restricted to the occurrence and duration of drought during a particular year: a mixture of drought-driven and fuel-limited fire regimes operated, emphasizing the lagged effects of warm or moist periods on fire (27). Higher fire activity was related to wetter conditions in the last three years. This illustrates that fuel abundance is an important constraint on Euro-Mediterranean (and other) fire regimes, even when drought is the main driver.
With respect to human alterations of fire regimes, contrasting short-term impacts of changes in land-use and fire suppression policy have been found: more fires where fuel biomass is high as a result of land-use change, and less fires in fire-prone ecosystems due to more effective fire suppression (28). Indeed, in the Mediterranean lowlands, which are densely populated and highly susceptible to extreme fire weather due to strong winds (29), both winter and summer fire activity were strongly suppressed, suggesting a gradual increase in the efficiency of fire suppression policy (30).
Changing drought conditions linked to extreme wildfires in northern Mediterranean
During the summers of 2003 and 2016 large parts of Europe suffered from extreme wildfire events. Experts questioned how changing drought conditions might affect the activity of these types of fires. For Mediterranean France, for instance, several fires became particularly large and devastating during these summers despite growing efforts in fire management and suppression capacities implemented since the beginning of the 1990s (32).
“Press droughts” and “hot droughts”
Looking for answers, scientists analyzed the weather conditions associated with these extreme wildfires. They concluded that extreme wildfire events result from specific interactions between different types of drought and different types of weather conditions. Two types of droughts are particularly responsible for these extreme events: the so-called “press droughts” and “hot droughts” (31).
“Press droughts” are subtle but chronic reductions in water availability, driven by long-term (month to seasons) reductions in precipitation and/or warmer temperatures, which increase potential evapotranspiration and reduce soil moisture (33). In 2016, a long lasting “press drought” intensified wind-driven fires.
In contrast, “pulse droughts” (or “flash droughts”) (34) only last days to weeks but are extreme in magnitude. The impact of a pulse drought on a long lasting press drought may intensify the conditions into a “hot drought” (35). This happened in 2003 when a long lasting drought and a summer heat wave combined.
Major shifts in meteorological conditions
It is generally accepted that a warmer and drier climate will alter the frequency, intensity or severity of wildfires. The analysis of the circumstances leading to the extreme wildfires of 2003 and 2016 illustrated two major shifts in meteorological conditions (31).
The first shift is the way that wind-driven wildfires are modified by press droughts, as illustrated by the 2016 fire season. Clearly, long lasting and intense droughts can substantially desiccate the living vegetation, thus increasing the impact of wind-driven fires, and therefore increasing fire intensity and rate of spread, and reducing suppression opportunities.
The second shift is the way that a heat wave adds to a long lasting drought, like in the summer of 2003, thus increasing the vulnerability to extreme wildfires even in the absence of strong winds. Probably, these conditions lead to the fast desiccation of plants during hot droughts (36), thus increasing fire intensity that may overwhelm fire suppression policies.
Changing drought conditions call for changes in fire management
These types of extreme wildfire events reflect the expected impacts of climate change on droughts in the Mediterranean (37) and in other regions of the world (38). “Hot droughts” and “press droughts” can lead to fire weather conditions that have not been explored before and to a subsequent increased frequency of extreme wildfire events. The dryness of fuel is pivotal in the occurrence of these events. The authors question whether some possible changes in fire management could compensate for increasing droughts and its impact on wildfire behavior (31).
Fire policy that focuses on suppression only delays the inevitable, promising more dangerous and destructive future forest fires (3,45). In contrast, land management agencies could identify large firesheds (20,000 to 50,000 ha) where, under specified weather conditions, managed wildfire and large prescribed fire are allowed to burn, sometimes after strategic mechanical fuel treatments (3). Acknowledging diversity in fire ecology among forest types and preparing forests and people for larger and more frequent fires could help reduce detrimental consequences. New strategies to mitigate and adapt to increased fire are needed to sustain forest landscapes. The following strategies have been suggested (5):
- Landowners should follow “Firewise” guidelines (www.firewise.org/) for houses and other infrastructure. Increased development in fire-prone landscapes has increased suppression costs, exacerbated risk to human safety and infrastructure, and reduced management options. People living in these forests must be prepared rather than relying solely on fire departments. Some places may be so hazardous that building should be prevented, discouraged, or removed (e.g., by regulation or insurance and/or tax incentives).
- Fire managers should avoid trying to uniformly blacken wildfire landscapes through burnout and mop-up operations, especially in burn interiors. As wildfire sizes have grown in recent decades, direct attack has been replaced with indirect attack, where fire lines are placed some distance from the active fire front, and then the area between is intentionally burned, often with high-severity fire, to reduce fuel and create a wider fire barrier. Unburned or partially burned patches are critical refugia that aid postfire recovery in forests of all fire regimes and should be conserved whenever possible.
- Land managers could anticipate changes using models of species distribution and ecological processes and should consider using assisted migration (4).
- Strategies should be based on a forest’s historical fire regime; in forests with historically high-frequency, low- to moderate-severity fire regimes the resilient forest structure should be restored similar to historical patterns that survived during past high-fire periods (and those anticipated in the future).
- Forest restoration should be funded.
Estimates of potential increase in annual burned areas in Europe under a high-end scenario of climate change (the so-called A2-emissions scenario) show an increase of about 200% by 2090, compared with 2000 – 2008, when no adaptation measures are taken (16). With respect to these estimates, the potential effectiveness of two adaptation options was assessed: (1) fire prevention through fuel reduction via prescribed burnings, and (2) active response through better fire suppression. The results of the assessment indicate that application of prescribed burnings may keep the increase in annual burned areas in Europe below 50%. Improvements in fire suppression might reduce this impact even further; e.g. boosting the probability of putting out a fire within a day by 10% country wide would result in about 30% decrease in annual burned area for that particular country. Additional adaptation options, such as using agricultural fields as fire breaks and behavioural changes, can potentially reduce the size of burned areas in the future even further (16).
Adaptation strategies - paradigm change for Mediterranean-type regions
Wildfire risk is high in the summer in Mediterranean-type climate regions. These regions are distributed over five continents: Africa, Australia, Europe, North America, and South America. They share a strongly seasonal climate, with cool, wet winters that promote vegetation (fuel) growth, and hot, dry summers that enhance vegetation flammability (45).
In recent decades, millions of new inhabitants and homes have moved into the wildland-urban interface in these regions. Human alterations of landscapes, and warming and drying climates plus ignitions (most often anthropogenic) during periods of severe fire weather have led to an increased prevalence of extreme wildfire events that often result in very large burned areas and significant impacts on human lives and assets (46). The answer to this has been to increase expenditure on reactive fire suppression. This strong focus on fire suppression is destined to fail, scientists from all continents argue. Focus must shift to mitigation and adaptation instead (45). The perspective of these experts is summarized below.
Fire suppression ineffective at extreme wildfires
Most experts agree that global warming will increase fire danger and burned areas in Mediterranean-type climate regions (47). The impact of global warming will be further exacerbated by ongoing changes in land use and management that increase fuel loads and continuity (48). Land use changes include expansion of human settlements into fire-prone areas, introduction of and invasion by fire-promoting exotic species, establishment of large, poorly managed tree plantations of highly flammable species, and agricultural land abandonment as a consequence of rural depopulation, resulting in replacement by unmanaged vegetation (49). Together, these trends lead to an increase in the amount and connectivity of fuel at the landscape-level, as well as the expansion of wildland-urban interface and inter-mix areas.
It is mostly extreme weather (fire weather) that drives extreme wildfire events. Land cover type has little impact on the spreading of the fire (50), except where large-scale and sustained strategic fuel reduction activities are implemented (51). Under these extreme conditions, fire suppression is largely ineffective even in cases of massive resource deployment (52). This is due to a combination of factors including strong winds that preclude ground engagement and aerial support, simultaneity of ignitions, and fire intensity above extinction capacity (53).
Shortsighted policies lead to firefighting trap
By largely ignoring climate warming and landscape-scale buildup of fuels, policy makers have fallen into the so-called ‘firefighting trap’ (54). Most of the investment in fire management has been allocated to fire suppression. This has contributed to ongoing fuel accumulation and landscape-level fuel continuity, and thus, paradoxically, has exacerbated the problem: it is getting more difficult to suppress fires under extreme fire weather, leading to more severe and usually larger fires.
The wildfire suppression approach is shortsighted, the scientists argue in their perspective paper. It seeks to minimize burned area in the short-term, treats fire as delivering only negative impacts, and tends to react to public opinion with ever-greater investment in firefighting capacity. Also, post-fire management, when implemented, is not always oriented to fire hazard mitigation in the medium/long-term. As a result, current land use and policy settings will likely result, in the long run, in larger burned areas and/or a greater share of total burned area being accounted for by the largest, and most intense fires (50,54), exacerbating both ecological and socio-economic impacts.
Aim at reducing damage, rather than area burned
The scientists stress that no amount of investment in suppression will prevent extreme wildfire events (Adams and Attiwill, 2011), in particular if the climate of Mediterranean-type climate regions is to become warmer and wetter, driving productivity and thus flammable biomass (55). Focusing on reducing area of land burned in any given year will merely postpone them (56). In fact, extreme fire weather and landscape-scale fuel hazard may confluence and generate fires of extraordinary intensity, seriously threatening lives, property and ecosystems.
According to these experts, the only alternative is to aim for reduced fire severity across large areas and in key locations, to minimize negative impacts to society, ecosystems and their services. Focus must shift from targets emphasizing reductions in area burned to targets more closely related to reducing fire negative impacts, including human lives lost, direct economic losses, soil erosion, water and air quality, carbon emissions, and biodiversity impacts. They propose that governments develop and implement a policy based on two key elements: promoting less vulnerable and more fire-resilient landscapes; and minimizing risk for humans and infrastructure.
Targeting the reduction of the amount and connectivity (landscape design) of fuels would reduce fire growth rate, increase the potential for fire suppression, and mitigate fire damage. Afforestation, reforestation and forest management should incorporate these aims, for instance by including species selection considering flammability. Agricultural policies should be better aligned with forest and fire policy, particularly in the Mediterranean Basin where maintaining farmland areas surrounding villages can help avoid vegetation encroachment around assets. Under controlled conditions, prescribed burning or fuel reduction burning is a very cost-effective fuel treatment to reduce hazard and suppress fire (57).
Interaction with local population
Also, programs may be set up to promote the removal of fuels by local communities. In addition, post-fire management provides a window of opportunity to implement large-scale and socially acceptable changes in forest and landscape planning that can create more fire-resilient and less flammable landscapes. In areas undergoing agricultural land abandonment, encroachment of highly flammable vegetation and tree plantations around rural settlements ought to be contained (45).
For the local population, community preparedness is a key component of a policy targeting reduced damage. This includes the definition of ‘stay-or-go’ policies, safe escape routes, and the engagement of local communities in the design and planning of mitigation actions (58).
Reducing anthropogenic fire ignitions remains an important component of all fire management strategies although, if not matched with the management of fuels, it will contribute to the firefighting trap (45).
A policy shift from suppression to mitigation and adaptation
The scientists stress that fire suppression must continue to play a key role in the protection of human lives and assets in Mediterranean-type climate regions. They also conclude, however, that extreme wildfire events will occur more often even in the face of escalating fire suppression expenditures. The focus on fire suppression must therefore shift to one on mitigation, prevention, and preparation. This may be difficult, they realize: fire suppression, when it works, has and immediate effect and is visible to the media, while fuel management is a much less visible investment on the long-term, and out of synchrony with electoral cycles. Still, they conclude, replying to each catastrophic fire season with ever increasing fire suppression expenditure, while disregarding mitigation and adaptation, will continue to be a major political mistake (45).
The references below are cited in full in a separate map 'References'. Please click here if you are looking for the full references for Europe.
- Attiwill and Binkley (2013); Moriondo et al. (2006); Flannigan et al. (2009), all in: Stephens et al. (2013)
- Attiwill and Binkley (2013), in: Stephens et al. (2013)
- North et al. (2012), in: Stephens et al. (2013)
- Iverson and McKenzie (2013), in: Stephens et al. (2013)
- Stephens et al. (2013)
- Pausas and Fernández-Muñoz (2012), in: IPCC (2014)
- Koutsias et al. (2012), in: IPCC (2014)
- Lavalle et al. (2009), in: IPCC (2014)
- Camia and Amatulli (2009); Carvalho et al. (2011); Hoinka et al. (2009); Koutsias et al. (2012); Salis et al. (2013), all in: IPCC (2014)
- Marques et al. (2011); Pausas and Fernández-Muñoz (2012); Fernandes et al. (2010); Koutsias et al. (2012), all in: IPCC (2014)
- Carvalho et al. (2011); Dury et al. (2011); Lindner et al. (2010); Vilén and Fernandes (2011), all in: IPCC (2014)
- Arca et al. (2012); Lung et al. (2012), both in: IPCC (2014)
- Pellizzaro et al. (2010), in: IPCC (2014)
- Dury et al. (2011), in: IPCC (2014)
- Rosan and Hammarlund (2007), in: IPCC (2014)
- Khabarov et al. (2016)
- Kean et al. (2016)
- Fréjaville and Curt (2017)
- Ganteaume and Jappiot (2013), in: Fréjaville and Curt (2017)
- Fréjaville et al. (2016), in: Fréjaville and Curt (2017)
- Batllori et al. (2013), in: Fréjaville and Curt (2017)
- Moreira et al. (2011), in: Fréjaville and Curt (2017)
- Moreira et al. (2011); Pausas and Fernández- Muñoz (2012), both in: Fréjaville and Curt (2017)
- Pezzatti et al. (2013); Moreno et a.l (2014), both in: Fréjaville and Curt (2017)
- Higuera et al. (2015), in: Fréjaville and Curt (2017)
- Curt et al. (2016), in: Fréjaville and Curt (2017)
- Keeley (2004); Pausas (2004); Meyn et al. (2007); Zumbrunnen et al. (2009); O’Donnell et al. Pausas (2011), all in: Fréjaville and Curt (2017)
- McWethy et al. (2013), in: Fréjaville and Curt (2017)
- Ruffault et al. (2017), in: Fréjaville and Curt (2017)
- Turco et al. (2014), in: Fréjaville and Curt (2017)
- Ruffault et al. (2018)
- Fox et al. (2015); Ruffault and Mouillot (2015); Curt and Frejaville (2018), all in: Ruffault et al. (2018)
- Hoover and Rogers (2016), in: Ruffault et al. (2018)
- Mo and Lettenmaier (2016), in: Ruffault et al. (2018)
- Overpeck (2013), in: Ruffault et al. (2018)
- Williams et al. (2012); Allen et al. (2015), both in: Ruffault et al. (2018)
- Sousa et al. (2011); Hoerling et al. (2012), both in: Ruffault et al. (2018)
- Mazdiyasni and Aghakouchak (2015), in: Ruffault et al. (2018)
- Turco et al. (2018)
- Amatulli et al. (2013); Migliavacca et al. (2013); Wu et al. (2015); Khabarov et al. (2016), all in: Turco et al. (2018)
- Bowman et al. (2017), in: Turco et al. (2018)
- Moritz et al. (2014), in: Turco et al. (2018)
- Abatzoglou et al. (2019)
- Greve and Seneviratne (2015); Gudmundsson and Seneviratne (2016), both in: Abatzoglou et al. (2019)
- Moreira et al. (2020)
- Bowman et al. (2017), in: Moreira et al. (2020)
- Bedia et al. (2015), in: Moreira et al. (2020)
- Pausas and Fernández-Muñoz (2012), in: Moreira et al. (2020)
- Pausas and Fernández-Muñoz (2012); Moreira et al. (2011); Gómez-González et al. (2017); Moreira and Pe’er (2018), all in: Moreira et al. (2020)
- Moreira et al. (2011), in: Moreira et al. (2020)
- Boer et al. (2009), in: Moreira et al. (2020)
- Brotons et al. (2013); Fernandes et al. (2016), both in: Moreira et al. (2020)
- Adams and Attiwill (2011), in: Moreira et al. (2020)
- Collins et al. (2013), in: Moreira et al. (2020)
- Batllori et al. (2013), in: Moreira et al. (2020)
- Stephens et al. (2013), in: Moreira et al. (2020)
- Fernandes et al. (2013), in: Moreira et al. (2020)
- Gill and Stephens (2009), in: Moreira et al. (2020)