Modelling wildfire occurrence at regional scale from land use/cover and climate change scenarios

Highlights

• A modelling framework for estimating future wildfire occurrence in LULC and climate change scenarios is described.

• LULC-derived interfaces and a combination of LFMC and seasonal climate-related variables were used as predictors.

• Expected changes will mainly increase wildfire occurrence with varied effects by analyzed region.

• Future regional wildfire predictions are useful to establish mitigation measures and for planning fire prevention actions.

Abstract

Wildfire occurrence is expected to increase in future climate and Land Use Land Cover (LULC) change scenarios, especially in vulnerable areas as the European Mediterranean Basin. In this study future probability of wildfire occurrence was estimated for a 20-year time period (2041–2060, centered on 2050) by applying a statistically-based regression model using LULC-derived contact areas with the forest cover (interfaces) as proxy for the human-related factor and a combination of Live Fuel Moisture Content and seasonal climate-related variables as predictors. Future wildfire occurrence was mapped under RCP 8.5 high emissions scenario in four Spanish regions with heterogeneous socioeconomic, LULC and natural fire-related characteristics at 1 km² target spatial resolution. Results showed increased wildfire probability in ∼19–73% of 1 km² cells, observing regional differences in the variable effects. This approach could be applied to other spatial scales offering tools for planning and management actions and to obtain different possible future scenarios.

Introduction

Nowadays wildfires are one of the significant threats to forested areas worldwide, and the European Mediterranean Basin is one of the more susceptible areas to fire episodes, reporting more than 85% of the total burned area in Europe (San-Miguel-Ayanz et al., 2012). Wildfires, however, are complex phenomena involving multiple factors mediate, e.g. fuel availability, moisture conditions, natural and human ignitions, meteorological/climate drivers (Gudmundsson et al., 2014) and management decisions (Hély et al., 2001), which also operate at different spatial and temporal scales (Bedia et al., 2015).

Reported changes in the Mediterranean region regarding the use of land and climate (Spinoni et al., 2020) are affecting the fire cycle (Pausas and Fernández-Muñoz, 2012), increasing the frequency and severity of wildland fires (Moreno et al., 2013), and threatening ecosystem stability, the provision of services, habitat and biodiversity conservation, landscape value and aesthetics, as well as property and human lives. These changes are expected to become more intense in the coming century (Syphard et al., 2019), increasing their effects on fires and the consequential impacts on human communities worldwide. Spain is representative of these changes within Mediterranean Europe (Stellmes et al., 2013), as lengthened, and longer fire weather seasons have become more frequent (Jolly et al., 2015). Understanding the past relationships among land use land cover (LULC) changes, climate and wildfire occurrence will allow for the prediction of future impacts and the evaluation of vulnerabilities, which will serve as input for management and policy actions (Gallardo et al., 2015). Resulting relationships will depend on the type and force of driving factors and their importance in different regions.

The European Mediterranean Basin has experienced profound LULC changes derived from human activities (Geri et al., 2010). In the last century, LULC changes in rural areas of Southern European countries were first linked to the intense abandonment of the countryside (1950s–1960s). Then, there was a shift to a new agricultural and mechanized system (1980s), followed by urban expansion in forested areas and increased recreational activities and the consequential intensification of the pressure on natural zones (2000s) (Vilar et al., 2016a). Such abandonment led to exceptional fuel accumulation due to natural reforestation processes (Geri et al., 2010) and urban pressure due to the increase in contact areas between forest and urban constructions (the so-called Wildland Urban Interface or WUI), all of which triggered an increase in wildfire risk. Besides, LULC also creates an impact through the ignition of fires, with more than 80% of fires in this area being linked to human activities and the result of negligence, accidents or acts of arson (Ganteaume et al., 2013), e.g., the use of fire to control herbaceous vegetation for cattle grazing or to clean brushwood in crops (lightning causes ∼5% in average of the known fires in this basin). Given the wide range of the human activities effects, several studies include the contact areas between forest and other covers, the so-called interfaces, such as WUI (Vilar et al., 2016a; Chas-Amil et al., 2013; Modugno et al., 2016; Lampin-Maillet et al., 2011), agricultural-forest interface (Gallardo et al., 2015; Martínez et al., 2009; Rodrigues et al., 2016) and grassland-forest interface (Rodrigues, 2014; Vilar et al., 2019), among others as human drivers of wildfire occurrence. Assuming that socioeconomic changes are likely to continue, further changes in LULC are also expected. Projecting the amount of LULC change and the location thereof will allow for future LULC derived interfaces to be obtained and for the human factor of the wildfire occurrence be represented (Gallardo et al., 2015).

On the other hand, the climate role in wildfires is mainly linked to the control of vegetation characteristics and status (Westerling et al., 2011). Pre-fire-season weather conditions have been proved to have a strong influence on the ignition and propagation of large fires because of their effect on fuel load and flammability (Urbieta et al., 2015). Moreover, the fuel moisture content (FMC), defined as the mass of water contained within vegetation per dry mass, is a critical variable affecting fire interactions with fuel (Yebra et al., 2013), and might affect both fuel ignition and fire spread rate (Viegas et al., 1992; Rossa et al., 2017). As FMC increases, the flammability of fuels tends to decrease, as more energy is needed to evaporate water before burning organic tissues (Argañaraz et al., 2018). FMC is usually separated into live (LFMC) and dead fuels (DFMC) (Chuvieco et al., 2004). Most operational fire danger rating systems include the estimation of DFMC, those lying on the forest floor (leaves, branches and debris) (Camia et al., 2003). Still, the estimation of LFMC is included less often. Less significant relations between fire spread or intensity were found in experimental data field analysis for a shrub or conifer forest (Fernandes and Cruz, 2012). Among other reasons, this is because LFMC is the result of complex interactions between previous and concurrent weather and the varied biological mechanisms that influence water content and dry matter accumulation (Jolly et al., 2014; Turner, 1981).

Climate change trends in Southern Europe are expected to lead to increased temperature, a greater number of heatwaves and dryer days (Cramer et al., 2008) and a decreased summer precipitation (Kovats et al., 2014). Modelling studies predict that this will lead to an increase in fire activity (Sousa et al., 2015), the number of large fires (Vázquez de la Cueva et al., 2012) and the burnt area (Amatulli et al., 2013; Turco et al., 2018). Dupuy et al. (2020) recently reviewed 23 studies that projected fire danger indices or fire activity (number of fires, size and burnt area) and at modelled climate-fire relationships in the European context at local, regional or continental scale. Results showed a relative increase (2–4% per decade) in mean seasonal fire danger under pessimistic climate change scenarios in the Mediterranean regions. Burnt areas are projected to increase everywhere in Southern Europe at a rate of 15–25% per decade (Dupuy et al., 2020).

Several simulation scenario studies have combined factors on humans, topography, vegetation and FMC along with climate change conditions as drivers of future wildfires in fire-prone areas worldwide. Examples of variables included in the modelling processes include, housing density (Westerling et al., 2011), distance to populated places (Liu et al., 2012), land use effects (Syphard et al., 2018), distance to roads (Syphard et al., 2019), road density (Liu et al., 2012), WUI and other land cover interfaces (Gallardo et al., 2015), aspect and slope (Westerling et al., 2011), vegetation type (Syphard et al., 2018) and fine fuel moisture content (Liu et al., 2012). However, in the European context, Dupuy et al. (2020) claimed that one of the sources of uncertainty in future estimations was the lack of information on the influence of human factors on climate-fire relationships. Moreover they affirmed that the influence of FMC should be taken into account and constitutes a possible source of bias for future predictions.

The main aim of this work was to develop an integrated modelling framework at 1 km² target resolution to better understand the importance of climate, LFMC and human factors on future spatial and temporal characteristics of wildfire occurrence in different Southern European regions. To that end, the variable effect and distribution of current and future projected probability of wildfires were analyzed in four areas of Spain. These regions were selected as representative of LULC changes and climate conditions in Southern Europe and have different socioeconomic, biophysical and wildfire characteristics. The specific objectives were first, to calibrate statistical-based regression models for wildfire occurrence combining climate, LFMC, topography and LULC interfaces. Secondly, the regression models were applied to the projected LULC and climate variables obtaining the future wildfire probability. The three main research questions addressed are, (1) would wildfire probability increase or decrease in relation to LULC and climate changes? and how?; (2) which variables would be more influential in the different regions?; and (3) could this modelling framework be applied to other scales and areas for planning and management actions?