Fire weather risk assessment under climate change using a dynamical downscaling approach

https://doi.org/10.1016/j.envsoft.2011.03.012Get rights and content

Abstract

Results from general circulation models suggest that the increase of forest fire activity at the global scale will be one of the impacts of climate change. As attention shifts to regional climate further spatial resolution is needed to handle the forcings and circulations that occur at smaller scales. One of the available techniques to assess the impact of climate change on fire activity at the regional scale is the dynamical downscaling between global climate models and regional models.

In the present work, the impact of climate change on fire danger at the regional scale was examined by means of dynamical downscaling between a general circulation model (MUGCM) and a regional meteorological model (MM5). A Southern European country, Portugal, was selected as case-study since general circulation models predict significant surface air temperature increases over Southern and Mediterranean Europe.

Present and future climates, centred in 1990 and 2050, respectively, were obtained using daily data previously simulated by MUGCM. Climate change signals on temperature and precipitation derived from the MUGCM ensemble simulations were analysed using spatial averages over the Iberian Peninsula and cluster analysis applied over Portugal. For the Iberian Peninsula, a positive trend for temperature for all seasons, with higher variability for the winter months, was obtained. Over Portugal, for the future climate, the average winter temperature is expected to be higher. Precipitation increases are simulated for the end of autumn/beginning of winter, and negative changes are expected for the end of winter/beginning of spring and beginning of summer. The cluster analysis revealed important temporal changes on the meteorological variables which may be relevant for fire management planning, namely a longer fire season over Portugal is expected.

The spatial refinement of the projected climate change impacts on the fire weather risk over Portugal was performed through numerical downscaling between MUGCM and MM5. The MM5 outputs, at 10 km resolution, were used to estimate the Canadian Fire Weather Index (FWI) System components. Results show higher FWI values in the beginning of summer for the 2050 scenario. An increase of the maximum values of the Drought Code (DC) in the inner part of Portugal was also detected. An increase in the total area burned is anticipated, with the consequent increase of pollutants emissions.

Introduction

Climate and weather are of extreme importance for the ignition and propagation of forest fires. The climate (regarded as the slowly varying aspects of the atmosphere – hydrosphere – land surface system) determines the total quantity of vegetation available for combustion, and duration and degree of severity of the forest fire season. On the other hand, weather conditions (the state of the atmosphere at a given time and place) regulate the moisture content of the dead biomass, and consequently its flammability potential.

Forest fires ignition and spread are sensitive to a number of different interactions within an ecosystem, such as weather, fuel load/type and topography. Hence, weather and climate play a crucial role in determining the fire regime of an area; in return, the fire regime is very sensitive to changes in climate (Pyne et al., 1996, Viegas et al., 1999, Skinner et al., 2000, Kunkel, 2001, Pereira et al., 2005). In Canada, significant relationships were established between historical area burned and weather (Harrington et al., 1983, Flannigan and Harrington, 1988). Other studies also pointed out that weather and climate are the most important natural factor influencing forest fires (Stocks and Street, 1983, Flannigan and Wotton, 2001, Hely et al., 2001). Considering these interactions, a question is raised: how will forest fire activity be in the future climate?

Since the 80s, when global climate modelling was in its early stages, a few studies were conducted in order to establish the links between future climate change and forest fire occurrences (Beer et al., 1988, Street, 1989). Based on global climate models results, the Intergovernmental Panel on Climate Change (IPCC) report (Fischlin et al., 2007) suggests that, with global warming, forest fires frequency will increase all over the world. In particular over Europe, General Circulation Models (GCMs) predict increases in surface air temperatures, especially in the southern part and the Mediterranean. This has the potential to create a drier forest at both the surface and in the organic layers of the forest soil through increased potential evaporation, especially during the summer months (Boer et al., 2000).

A recent study integrating vegetation response to climate change and surface hydrological processes (Alo and Wang, 2008), used offline simulations with a land-surface model to explore the future response to elevated carbon dioxide (CO2) concentrations. The authors demonstrated that CO2-induced changes in potential vegetation structure substantially influences the surface hydrological processes, concluding that the soil moisture simulated by different GCMs varies when considering changes on vegetation structure and that the hydrological cycle may, indeed, intensify.

Assessments of the potential impacts of climate change on fire weather risk in the forests of Canada and United States (Flannigan and Van Wagner, 1991, Stocks et al., 1998, Wotton et al., 1998, Flannigan et al., 1998, Flannigan et al., 2000) have used GCMs outputs to project fire severity using, for example, components of the Canadian Fire Weather Index (FWI) System (defined in section 2.2). Results have shown increasingly severe fire weather across most of the western boreal forest of Canada and United States. In addition to rises in seasonal means of fire severity indices, these studies also predict enhancements in the frequency of occurrence of extreme fire severity in specific areas (Wotton et al., 2003) and increase in the fire season length (Wotton and Flannigan, 1993).

Research with respect to climate simulation and prediction has attracted considerable efforts throughout the last 30 years with global aspects clearly dominating. However, it is the regional and the local climate that is of central importance to societies and to the biosphere (Grell et al., 2000). The horizontal atmospheric resolution of present day GCMs is still relatively coarse, and regional climate is often affected by forcings and circulations that occur at smaller scales (Giorgi and Mearns, 1991). As a result, GCMs cannot explicitly capture the fine scale structure that characterizes climate variables in many regions of the world and which is needed for many impact assessment studies. Examples of fine scale phenomena are mesoscale processes that contribute for intense precipitation events, like those promoted by vertical upward motions and coupling between regional circulation and convection; local and extreme winds, as well as extreme temperature values, due to topographic effects and/or short term variability.

In order to better simulate the fine scale atmospheric processes it is necessary to use mesoscale models, also named regional or limited area models. This type of models, with time scales less than 1 h and spatial resolution of a few kilometres, have the possibility to resolve land surface features such as strong gradients on topographic and land-use features that may be important for land-atmospheric fluxes calculations (Bader et al., 2008). Topography, land-use and land-water mask, are examples of important static fields that influence regional to local weather. Mesoscale models need lateral boundary conditions obtained either from observations or from global climate models. The method by which a mesoscale model is forced in its lateral boundaries, and eventually initialized, by a global climate model is called dynamical downscaling (Wilby et al., 2002, Christensen et al., 2007a, Thatcher and Hurley, 2010).

Benefits from this approach over European regions are explored in Bojariu and Giorgi (2005). The authors show the importance of having a better description of topographic features in the presence of a realistic account of the atmospheric processes when detecting the Northern Atlantic Oscillation (NAO) in climate simulations. This study compares the obtained correlations between observed pressure and the GCM model results, with a horizontal resolution of 1.25° latitude by 1.875° longitude, and also the regional climate simulation (forced with the GCM outputs) results obtained with a horizontal resolution of 50 km over several regions of Europe, namely over the Iberian Peninsula. When using dynamical downscaling the authors were able to simulate the large scale NAO variability taking into account the fine scale processes. Moreover, they were able to determine the importance of topography on the interaction between the large scale NAO and precipitation.

During summer, Europe experiences a large number of forest fires which may cause enormous losses in terms of environmental damage and even human lives. Most of the fires take place in the Mediterranean region, with more than 95% of its area burned (EC, 2003). Consequently, governments currently allocate a part of their national budget to prevent, fight and remediate forest fires.

The Mediterranean is located in the subtropics and, as such, has been experiencing a warming and drying trend in the last decades (Trenberth et al., 2007). Also, climate projections from the latest IPCC report (Christensen et al., 2007b), resulting from a regional multi-model application, show that these trends, which are stronger and of opposite sign of those in the mid-latitudes in the European region, should continue and even intensify in the future. Christensen et al. (2007b) present results from climate projections of an ensemble of 21 models; regarding temperature and precipitation responses over Europe, Table 1 shows the median of annual average and warm season (June, July and August) temperature anomalies and precipitation responses (in percentage) for Northern Europe (NE) and Southern Europe and Mediterranean (SEM).

Warmer and drier conditions will increase the forest fire risk in the region (Moriondo et al., 2006). Portugal is one of the European countries most affected by forest fires, mainly during summer, which is characterised by hot and dry weather (EC, 2005). In Portugal, the peak season of wildfires takes place between June and September, with 93% of the annual area burned (Hoinka et al., 2009). The number of fire occurrences in the months of June (8%), July (22%), August (32%), and September (20%) represent 82% of the yearly total (Carvalho et al., 2008).

In Pausas (2004) the link between forest fire occurrences and climate variables is analysed for the Valencia region in Spain, concluding that summer rainfall is an important factor for determining the amount of area burned. The author also concludes that fire ignitions may be determined by human factors, while some of the variability in the annual area burned is explained by climatic parameters. In Portugal, Viegas et al. (1992) and Viegas and Viegas (1994) established a clear dependency between the annual area burned and the total rainfall from May to September. The authors also stressed that the rainfall in the beginning of the fire season, namely in June, has a marked influence in the reduction of the area burned. More recently, Carvalho et al. (2008) performed regressions between the area burned and fire occurrence with the meteorological and the Canadian FWI variables for the period between 1980 and 2004 for twelve districts across Portugal. Their regression approach explained 61–80% of the variance in area burned and 48–77% of the variance in the fire occurrence, depending on location (p < 0.0001).

The main objective of the present study is to investigate the future impacts of climate change scenarios on fire weather risk at the regional scale. A dynamical downscaling approach was applied from the GCM outputs to the regional scale model to better capture the effects of local and regional forcing in an area of complex topography, such as Portugal. In addition, the comparison between estimated and measured fire weather risk indexes was also performed. The paper is organized as follows: section 2 describes the data and methodology used; the results and discussion are examined in section 3; and section 4 provides the main conclusions.

Section snippets

Global and regional climate simulations

Present and future climates (centred around 1990 and 2050 respectively) over Portugal were modelled by dynamical downscaling using the data simulated by the Melbourne University General Circulation Model (MUGCM) (Simmonds and Lynch, 1992) which were specified as initial and lateral boundary conditions to the Fifth-Generation Mesoscale Model (MM5) (Dudhia, 1993). The current fuel map data were used for the present and future climates and hence the potential interactions between

Climate simulation analysis

Hereafter, seasons are defined for the Northern Hemisphere as such: summer (June, July and August), autumn (September, October and November), winter (December, January and February) and spring (March, April and May).

MUGCM’s projected changes for the study region between the 1990 and 2050 climate are in agreement with those obtained by other models under similar global emissions scenarios (global climate model results for monthly mean temperature and precipitation may be consulted, for the

Conclusions

The Canadian Forest Fire Weather Index System, currently applied by the Portuguese authorities during the fire season, was selected for the evaluation of the impact of climate change on fire weather risk over Portugal.

The statistical analysis of serial data generated by the general circulation model for Portugal points towards an increase in the fire weather risk due to higher temperatures and lower precipitation. The cluster analysis applied to monthly mean temperature and monthly total

Acknowledgements

The authors wish to thank the three reviewers for their constructive comments on the manuscript. The authors are grateful to the Portuguese Foundation for Science and Technology and the European Social Fund, for the Structural Funds period between 2000–2006, in the scope of QUIMERA Project (POCTI/34346/CTA/2000) and for the PhD grants of H. Martins (SFRH/BD/13581/2003) and A. Carvalho (SFRH/BD/10882/2002). The authors are also grateful to the Network of Excellence ACCENT (GOCE-CT-2004-505337).

References (71)

  • D.C. Bader et al.

    Climate Models: An Assessment of Strengths and Limitations. U.S. Climate Change Science Program

    (2008)
  • T. Beer et al.

    Australian bushfire danger under changing climatic regimes

  • G.J. Boer et al.

    An Intercomparison of the Climates Simulated by 14 Atmospheric General Circulation Models

    (1991)
  • G.J. Boer et al.

    Transient climate change simulation with greenhouse gas and aerosol forcing: projected climate to the twenty-first century

    Clim. Dynam.

    (2000)
  • R. Bojariu et al.

    The North Atlantic Oscillation signal in a regional climate simulation for the European region

    Tellus

    (2005)
  • A. Carvalho et al.

    Fire activity in Portugal and its relationship to weather and the Canadian Fire Weather Index System

    Int. J. Wildland Fire

    (2008)
  • J.H. Christensen et al.

    Evaluating the performance and utility of regional climate models: the PRUDENCE project

    Clim. Change

    (2007)
  • J.H. Christensen et al.

    Regional climate projections

  • M.G. Cruz

    Descrição do Sistema Canadiano de Indexação do Perigo de Incêndio (Description of the Canadian Fire Weather Index System)

    (2000)
  • J. Dudhia

    A nonhydrostatic version of the Penn State – NCAR Mesoscale Model: validation tests and simulation of an Atlantic cyclone and cold front

    Mon. Weather Rev.

    (1993)
  • Forest Fires in Europe: 2002 Fire Campaign. Directorate-General Joint Research Centre, Directorate-General Environment

    (2003)
  • Forest Fires in Europe 2004. Directorate-General Joint Research Centre, Directorate-General Environment

    (2005)
  • A. Fischlin et al.

    Ecosystems, their properties, goods, and services

  • M.D. Flannigan et al.

    A study of the relation of meteorological variables to monthly provincial area burned by wildfire in Canada 1953–80

    J. Appl. Meteorol.

    (1988)
  • M.D. Flannigan et al.

    Climate change and wildfire in Canada

    Can. J. For. Res.

    (1991)
  • M.D. Flannigan et al.

    Climate, weather and area burned

  • M.D. Flannigan et al.

    Future wildfire in circumboreal forests in relation to global warming

    J. Veg. Sci.

    (1998)
  • F. Giorgi et al.

    Approaches to the simulation of regional climate change: a review

    Rev. Geophys.

    (1991)
  • R. Gnanadesikan

    Methods for Statistical Data Analysis of Multivariate Observations

    (1997)
  • G.A. Grell et al.

    Nonhydrostatic climate simulations of precipitation over complex terrain

    J. Geophys. Res.

    (2000)
  • J.J. Hack et al.

    Description of the NCAR community climate model (CCM2)

    (1993)
  • J.B. Harrington et al.

    A Study of the Relation of Components of the Fire Weather Index to Monthly Provincial Area Burned by Wildfire in Canada 1953–80

    (1983)
  • C. Hely et al.

    Role of vegetation and weather on fire behavior in the Canadian Mixed wood boreal forest using two fire behavior prediction systems

    Can. J. For. Res.

    (2001)
  • A. Henderson-Sellers et al.

    Using stable water isotopes to evaluate basin-scale simulations of surface water budgets

    J. Hydrometeorol.

    (2004)
  • K. Hoinka et al.

    Regional-scale weather patterns and wildland fires in Central Portugal

    Int. J. Wildland Fire

    (2009)
  • Cited by (46)

    • Assessing impacts of future climate change on extreme fire weather and pyro-regions in Iberian Peninsula

      2021, Science of the Total Environment
      Citation Excerpt :

      The close relationship between weather and wildfires (Pereira et al., 2005; Trigo et al., 2016; Amraoui et al., 2015; Vieira et al., 2020) suggests that climate change can impact future fire risk and burnt area (Pereira et al., 2013; Sousa et al., 2015). For the Iberian Peninsula, the literature suggests increasing fire risk for future climate conditions (Carvalho et al., 2011; Pereira et al., 2013; Sousa et al., 2015; Pérez-Sánchez et al., 2019). Meteorological parameters and fire indices, e.g., of the Canadian Forest Fire Weather Index System (CFFWIS) have been used to model current and future fire activity and to help forest and fire managers as well as national authorities in prevention and suppression activities all over the world (Li et al., 2008; Flannigan et al., 2013, 2016).

    • Uncovering the perception regarding wildfires of residents with different characteristics

      2020, International Journal of Disaster Risk Reduction
      Citation Excerpt :

      “Training” and “Prevention” are the predominant categories for most age groups, whereas “Fuel management” and “Community involvement” appear as options in people specifically between 36 and 49 years old, and in those with university education. In view of the expected increase in extreme events due to climate change [65–68] and the subsequent inability of suppression resources to tackle extremely large fires [61,69], community involvement and preparedness regarding wildfires become even more important. The associations found between individual characteristics and experience with wildfire perception and knowledge were based on a simple approach, using few and straightforward possibilities of response for most questions.

    • A system for airport weather forecasting based on circular regression trees

      2018, Environmental Modelling and Software
      Citation Excerpt :

      The need for higher resolution forecasts has driven numerous methodologies to generate more detailed outputs, which is known as downscaling. Dynamic downscaling uses the output of a coarser model as the initial condition of a higher resolution local model, which better resolves sub-grid processes and topography (Carvalho et al., 2011). Another approach is statistical downscaling, where historical observed data are used to enhance the output of a numerical model.

    View all citing articles on Scopus
    1

    Present address: Centre for Environmental and Sustainable Research, CENSE-DCEA-FCT/UNL Campus de Caparica, 2829-516 Caparica, Portugal.

    View full text