You are here

Article Text

Article menu

Download PDFPDF

Environment
Original article
Forest fires are associated with elevated mortality in a dense urban setting
  1. Antonis Analitis,
  2. Ioannis Georgiadis,
  3. Klea Katsouyanni
  1. Department of Hygiene, Epidemiology and Medical Statistics, Medical School, University of Athens, Athens, Greece
  1. Correspondence to Antonis Analitis, Department of Hygiene, Epidemiology and Medical Statistics, Medical School, University of Athens, 75, Mikras Asias Street, 115 27 Athens, Greece; aanalit{at}med.uoa.gr

Abstract

Objectives The climate and vegetation of the greater Athens area (population over three million) make forest fires a real threat to the environment during the summer. A few studies have reported the adverse health effects of forest fires, mainly using morbidity outcomes. The authors investigated the short-term effects of forest fires on non-accidental mortality in the population of Athens, Greece, during 1998–2004.

Methods The authors used generalised additive models to investigate the effect of forest fires on daily mortality, adjusting for time trend and meteorological variables, taking into account air pollution as measured from fixed monitors. Forest fires were classified by size according to the area burnt.

Results Small fires do not have an effect on mortality. Medium sized fires are associated with an increase of 4.9% (95% CI 0.3% to 9.6%) in the daily total number of deaths, 6.0% (95% CI −0.3% to 12.6%) in the number of cardiovascular deaths and 16.2% (95% CI 1.3% to 33.4%) in the number of respiratory deaths. Cardiovascular effects are larger in those aged <75 years, while respiratory effects are larger in older people. The corresponding effects of the one large fire are: 49.7% (95% CI 37.2% to 63.4%), 60.6% (95% CI 43.1% to 80.3%) and 92.0% (95% CI 47.5% to 150.0%). These effects cannot be completely explained by an increase in ambient particle concentrations.

Conclusion Forest fires have an immediate effect on mortality, not associated with accidental deaths, which is a significant public health problem, especially if the fire occurs near a densely populated area.

  • Forest fires
  • mortality
  • cardiovascular
  • respiratory
  • ambient particles
  • Greece
  • epidemiology
  • climate
  • mortality studies
  • time series study
  • environment

https://doi.org/10.1136/oem.2010.064238

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    What this paper adds

    • Forest fires, an important threat to the environment and to neighbouring populations, are predicted to increase in frequency due to climate change.

    • A few studies have reported the adverse health effects of forest fires, mainly using morbidity outcomes.

    • This study adds evidence that forest fires are associated with important short-term elevations in non-accidental cause-specific mortality.

    • The effect on mortality is proportional to the size of the fire and greater for respiratory causes.

    • The results of this study indicate considerable health effects of forest fires occurring near densely inhabited areas, which need to be addressed by designing appropriate emergency reaction plans.

    Introduction

    The climate of Greece is characterised by hot and dry summers.1 More than three million people live in the greater Athens area (Hellenic Statistical Authority, Census 2001),2 a basin in which the proportion of urban green area per inhabitant is small (about 2 m2).3 Hence, the importance of the peri-urban green areas on the surrounding mountains should always be considered. The heat and dryness, together with the prevailing very flammable pine tree vegetation, make forest fires a real threat during the summer. The increase in temperature observed in recent years, predicted to continue due to climate change, is expected to aggravate the forest fire problem.4

    Apart from accidents occurring during the fires, which are usually counted and publicised by the media, it is plausible to hypothesise that there may be other acute effects on health attributed to the high particle exposures during the fires, to collective stress or to other unidentified factors.5–8

    In other locations such as in North America, Australia and Southeast Asia with similar forest fire problems, an increase in hospital visits for respiratory disease exacerbation and in the frequency of respiratory symptoms has been reported.9–13 Only two studies have evaluated the impact on mortality with different approaches and rather diverging results.8 14 One of the studies found effects of particulate matter on mortality while the other reports no such clear evidence.8

    In this paper, we investigate the immediate effects of forest fires on the total and cause-specific mortality excluding accidental deaths in the population of the greater Athens area during the time period from 1998 to 2004.

    Materials and methods

    Study area

    Figure 1 shows the study area. The black line denotes the boundary of the greater Athens area covering about 400 km2. The areas within the Attica Prefecture where fires occur (to the east, northeast and northwest of the greater Athens area) are indicated. The location of the four fixed site air pollution monitors is also shown.

    Figure 1
    Figure 1

    Map of the study area indicating the greater Athens area, the location of the large fire and the usual locations of the smaller fires as well as the four fixed site monitors for black smoke during the study period.

    Data

    The daily number of deaths by age group (all ages, <75 and ≥75 years) was obtained from the Hellenic Statistical Authority for all natural causes (ICD9<800), cardiovascular causes (ICD9: 390–459) and respiratory causes (ICD9: 460–519).

    The dates of the forest fires and the area burnt around the greater Athens area were obtained by request for the years 1998–2004 from the Fire Service of Greece. Forest fires that burned more than 10 000 m2 were considered. The data are complete because the Fire Service is called upon when a fire occurs anywhere in the area. In table 1 the dates, location, duration of each fire and the area burnt are shown.

    View this table:
    Table 1

    Number of fires (n), area burnt (×1000 m2) and duration (days) by size of fire and year of study in the greater Athens area

    During the entire study period, the only index of ambient particulate matter measured routinely in Athens was black smoke (BS), which is based on reflectometry measurements subsequently converted to mass concentrations. BS represents black particles with an aerodynamic diameter <4 μm.15 The fixed site air pollution monitoring network in Athens is operated by the Ministry of Environment, Energy and Climate Change.16 We obtained the daily measurements by request from the Ministry and during the study period there were five sites monitoring BS. We only used stations with less than 25% of missing values over the whole period; thus, one station was excluded. The remaining missing values were replaced using the method described by Katsouyanni et al17 and an arithmetic average of all measurements was then calculated for each day. In the final time series of BS, there were only 4 days with missing BS levels. It should be noted that the fixed site monitors may not adequately represent particles generated from forest fires as they are placed to monitor mainly traffic generated pollution (see figure 1). Daily temperature (°C), humidity (%), wind speed (m/s) and prevailing wind direction (degrees) were obtained from the National Observatory of Athens.18 Meteorological series were complete. Only the period from May to September was analysed because this is the period in which more than 70% of the forest fires occur.

    Table 2 shows the mean daily number of deaths and the 24-hour mean levels of BS, temperature and humidity for the years 1998–2004.

    View this table:
    Table 2

    Descriptive characteristics for the daily number of deaths, black smoke concentrations, daily temperature and humidity on days with and without forest fires in Athens, Greece, 1998–2004

    Methods

    We applied generalised additive models19 (GAM) allowing for overdispersion (quasi-likelihood used) with the daily number of total natural and cause-specific deaths by age group used alternatively as outcome variables. Three dummy variables were included in the model for forest fires: one indicating the days during which a large forest fire occurred (one which burnt more than 30 000 000 m2), a second indicating days with medium size forest fires (the ones which burnt between 1 000 001 and 30 000 000 m2) and a third for days with small fires (that burnt between 10 000 and 1 000 000 m2). This categorisation was based on the distribution of the area burnt. If on a specific day there was more than one fire, the value of the area burnt that day was the sum of all areas burnt by simultaneous fires.

    To adjust for the potential confounding effects of temperature, we included a smooth function of temperature (average of lags 0–2) using penalised splines with 5 degrees of freedom.20 Additionally, we included a dummy variable to indicate heat wave days (to account for effects that may exist over those attributable to very high temperature). We followed the definition of heat wave used in the EuroHEAT project according to which a heat wave day occurs if the maximum apparent temperature21 and the minimum temperature exceed the 90th percentile of the same month's distribution over the whole study period and this exceedance lasts for more than two consecutive days.22 Relative humidity was also included as a covariate. Because wind speed and direction can be very important for the spread of smoke from wood fires, we also included wind speed and three dummy variables for wind direction (West: 46–135°; South: 136–225° and East: 226–315°, using the North: 316–45° as the reference category). To adjust for reported patterns of mortality17 we included six dummy variables for day of the week and one variable to indicate holidays. We used penalised splines to control for seasonal and long-term trends. The degrees of freedom were chosen using the PACF minimisation criterion proposed by Touloumi et al.23 The final model was of the form:

    Log(E[Yi])=β0+[β1Fs+β2Fm+β3Fl]+s(temp0−2)+s(time)+[confounders],

    where Yi represents the number of deaths on day i; Fs, Fm and Fl are the dummy variables for forest fires by size; s(temp) and s(time) are the smooth functions of the temperature and time trend, respectively. Confounders are day of the week, holiday, wind speed and direction.

    To investigate the possible additional lagged effect on mortality, we also fitted models assessing the effects of the fires during days with fires plus two days after they were extinguished. For these specific models, we preserved the same classification of forest fires in three categories, but the variables indicating days with fires also included the two days after extinction.

    Sensitivity analysis using dummy variables for each month of each year (a total of 34 variables) to control for long-term trends and natural cubic splines with 5 degrees of freedom to adjust for temperature was performed to check the robustness of the results.

    Models with BS measurements, average of lags 0 and 1, were also tested in order to assess how much of the effect of forest fires can be attributed to particulate pollution (at least the pollution measured by the fixed site monitoring system). Interaction terms between wind variables and forest fire variables as well as BS were tested but since they were not significant (p>0.20) they were removed from the model. All analyses were performed using the R software.24

    Results

    Table 3 shows the effects of forest fires on total and cause-specific mortality for all ages and by age groups without adjustment and after adjusting for long-term trends, temperature, humidity, wind speed and direction, week day and holidays. Smaller fires (those burning less than 1 000 001 m2) do not appear to have an effect on mortality. Medium sized fires are consistently associated with an increase in mortality (smaller for total, followed by cardiovascular and the highest for respiratory) but this does not reach the nominal level of statistical significance in some cases. The large fire of 1998, which burnt more than 70 000 000 m2, is associated with a very high increase in the daily number of deaths. Thus, this fire is associated with a 50% increase in the total daily number of deaths, 61% increase in the number of cardiovascular deaths (78% for those <75 years old and 55% for those ≥75 years old) and 92% increase in the daily number of respiratory deaths (72% for those <75 years old and 101% for those ≥75 years old).

    View this table:
    Table 3

    Crude and adjusted* per cent increase (95% CI) in the daily number of deaths in days with different sizes of forest fires by cause of death and age group for the years 1998–2004 in Athens, Greece

    The effects on total and cardiovascular mortality are higher during the days of the fires, while the lagged effects are larger for respiratory mortality (see online supplementary figures S1 and S2).

    We also additionally adjusted for the measured daily concentrations of BS. The effects of forest fires remain practically unchanged except for the effect of the large fire on respiratory deaths among those aged 75 years and older, which is reduced by 7% (see online supplementary table S1).

    The results from sensitivity analyses models using alternative methods to control for long-term trends produced very similar effect estimates.

    Discussion

    In this study, we investigated the acute effects of forest fires on mortality. We found that small fires appeared to have no effect, probably because their emissions are not high enough to affect a large proportion of the population, but medium- and large-sized fires have an effect on the daily number of deaths during the days of the fire. Consistently higher effects were observed for the one large fire. Cardiovascular mortality was affected more in the age group under 75 years while respiratory mortality was more strongly affected among older people (≥75 years). The effects appeared to be stronger and more persistent with time for respiratory mortality.

    Most studies investigating the effects of forest fires have assessed morbidity end points such as hospital admissions, physician or emergency room visits, asthma exacerbation or symptoms.6 9 11 They generally report increased effects either on forest fires days or per specific increase of particle concentration indicators (PM10 or PM2.5). One study from Australia reports no effect.25 We are aware of only two studies using mortality outcomes. Sastry14 investigated the effects of large forest fires (indeed among the largest events that have occurred worldwide) that occurred in Indonesia in 1997 on mortality in Malaysia. He was unable to study the effects on Indonesia itself due to the lack of adequate data. The exposure variables used were PM10 and visibility. In this study, the effects of smoke from long-range transportation were assessed and ‘high’ pollution days were defined using various PM10 thresholds above 10 μg/m3 and therefore the results are not strictly comparable to ours. Interestingly, Sastry also found cardiovascular mortality effects to be higher for age groups <75 years old and higher respiratory mortality effects for those ≥75 years old, in agreement with the results of the present study. Vedal and Dutton8 studied two air pollution episodes that occurred when smoke from a large wildfire was shifted over Denver, Colorado, due to the wind. They attempted to investigate the possible acute increase in mortality immediately following the particle concentration increase, without using modelling techniques to account for other potential variables that could affect mortality. Such obvious acute effects could not be clearly seen. However, the authors did not consider other pathways through which a fire may affect mortality and did not explore the possibilities provided by modelling the data in the ‘exposed’ and ‘control’ areas.

    Indeed, it is a characteristic of most reported studies that by design or by underlying hypothesis they test whether the effects of forest fires occur exclusively mediated through the elevation of ambient particle concentrations. In our particular situation, we find that not all the acute effects of forest fires on mortality may be explained in this way. In Athens, particulate air pollution during the study period was measured using the BS method, which measures particles of black colour smaller than 4 μm. Thus, it measures mainly carbonaceous, primary (emitted) particles. The fixed site monitors are placed in ‘urban background’ or ‘traffic’ sites within the city. Major forest fires in Greece often occur during days with strong winds, which facilitate the quick spreading of the fire and may also be associated with cleaning the usual air pollutants from the Athens basin.1 However, if smoke is continuously generated by a large forest fire to the northeast or west of the Athens basin, it is possible that particulate pollution may increase over inhabited areas nearer to the forest fires that are inadequately represented by the fixed site measurements. In our analysis, BS was not found to be particularly elevated during the days of the forest fires. This may be a result of inadequate monitoring or it may be an indication that other mechanisms operate to increase mortality during forest fires. The first hypothesis (that of inadequate monitoring) is supported by the high increase observed in respiratory mortality.

    Another mechanism that may operate during forest fires is the one associated with psychological stress as a predecessor of fatal cardiac death. The increase in cardiac mortality may be a result of this mechanism as well as a result of increased particle concentrations. The occurrence of a large-scale forest fire may be considered as a ‘natural experiment’ through which the population of an entire area is subjected to a stressful event of short duration, especially since, in addition to the experience and view of the fire and the smoke, the event attracts extreme publicity and attention from television channels. Similar increases in cardiac mortality have been reported before in relation to the occurrence of earthquakes.26–28 Thus, Trichopoulos et al26 reported that during the 3 days following a major earthquake in Athens, there was a 50% increase in fatal cardiac events. The excess deaths peaked during the third day and were more evident for those aged less than 70 years. Leor et al28 also reported a sharp increase in deaths related to cardiovascular causes (RR for atherosclerotic causes of death 2.6) on the day of a strong earthquake in Los Angeles County. Similarly, increased coronary deaths have been reported after terrorist attacks or war or even important sporting events.5 29 30 As Bhattacharyya and Steptoe report,5 the critical physiological mechanism responsible for stress-related death appears to be ventricular arrhythmia, which may triggered through the involvement of an intense psychological state burdening daily life and/or a proximate psychological event. Along the lines of attributing the elevation of mortality associated with forest fires to mechanisms other than increased ambient particle concentrations is the observation of Vedal and Dutton8 that mortality in one of the events in Denver started to increase before a measurable increase in PM10 was recorded. These mechanisms do not exclude a parallel effect of high exposure to elevated particle concentrations.

    In our study, we have not been able to compare the toxicity of particles originating from fires to that of particles generated from vehicle traffic. The latter source of particles is the most important contributor of particulate matter concentrations in Athens under usual conditions. An increase in the daily concentration of BS by 10 μg/m3 has been associated with a 1.5% increase in the total number of daily natural deaths.17 We were also unable to assess the increase in BS exposure of the population living nearer to the forest fires and therefore we cannot accurately estimate the number of deaths that may be attributed to exposure to particles. One way to investigate this issue further is to examine satellite pictures that show the movement of smoke across geographical areas. Unfortunately, these pictures lack the resolution of height needed to estimate particle concentrations at the level where people actually breathe (1.5–2 m height).

    Conclusion

    In conclusion, our results suggest that forest fires have important, indirect, adverse health effects. The number of forest fires is predicted to increase due to climate change4 and thus they may become an even more important public health problem that has not yet received adequate attention. Studies designed to better quantify the effect, enhance the understanding of mechanisms and estimate the health impact are needed.

    Acknowledgments

    The authors would like to thank the Fire Service of Greece for providing the data on forest fires.

    References

    1. ↵
      1. Kassomenos PA,
      2. Gryparis A,
      3. Katsouyanni K
      . On the association between daily mortality and air mass types in Athens, Greece during winter and summer. Int J Biometeorol 2007;51:315–22.
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      Hellenic Statistical Authority. Census. 2001. http://www.statistics.gr/portal/page/portal/ESYE (accessed 28 May 2010).
    3. ↵
      1. Damigos D,
      2. Kaliampakos D
      . Assessing the benefits of reclaiming urban quarries: a CVM analysis. Landsc Urban Plan 2003;64:249–58.
      OpenUrlCrossRefWeb of Science
    4. ↵
      1. Podur J,
      2. Wotton M
      . Will climate change overwhelm fire management capacity? Ecological Modelling 2010;221:1301–9.
      OpenUrlCrossRefWeb of Science
    5. ↵
      1. Bhattacharyya MR,
      2. Steptoe A
      . Emotional triggers of acute coronary syndromes: strength of evidence, biological processes, and clinical implications. Prog Cardiovasc Dis 2007;49:353–65.
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Moore D,
      2. Copes R,
      3. Fisk R,
      4. et al
      . Population health effects of air quality changes due to forest fires in British Columbia in 2003: estimates from physician-visit billing data. Can J Public Health 2006;97:105–8.
      OpenUrlPubMedWeb of Science
    7. ↵
      1. Naeher LP,
      2. Brauer M,
      3. Lipsett M,
      4. et al
      . Woodsmoke health effects: a review. Inhal Toxicol 2007;19:67–106.
      OpenUrlPubMedWeb of Science
    8. ↵
      1. Vedal S,
      2. Dutton SJ
      . Wildfire air pollution and daily mortality in a large urban area. Environ Res 2006;102:29–35.
      OpenUrlPubMed
    9. ↵
      1. Duclos P,
      2. Sanderson LM,
      3. Lipsett M
      . The 1987 forest fire disaster in California: assessment of emergency room visits. Arch Environ Health 1990;45:53–8.
      OpenUrlPubMedWeb of Science
    10. ↵
      1. Mott JA,
      2. Meyer P,
      3. Mannino D,
      4. et al
      . Wildland forest fire smoke: health effects and intervention evaluation, Hoopa, California, 1999. West J Med 2002;176:157–62.
      OpenUrlCrossRefPubMedWeb of Science
    11. ↵
      1. Sorenson B,
      2. Fuss M,
      3. Mulla Z,
      4. et al
      . Surveillance of morbidity during wildfires—Central Florida, 1998. MMWR Morb Mortal Wkly Rep 1999;48:78–9.
      OpenUrlPubMed
    12. ↵
      1. Delfino RJ,
      2. Brummel S,
      3. Wu J,
      4. et al
      . The relationship of respiratory and cardiovascular hospital admissions to the southern California wildfires of 2003. Occup Environ Med 2009;66:189–97.
      OpenUrlAbstract/FREE Full Text
    13. ↵
      1. Sutherland ER,
      2. Make BJ,
      3. Vedal S,
      4. et al
      . Wildfire smoke and respiratory symptoms in patients with chronic obstructive pulmonary disease. J Allergy Clin Immunol 2005;115:420–2.
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      1. Sastry N
      . Forest fires, air pollution, and mortality in Southeast Asia. Demography 2002;39:1–23.
      OpenUrlPubMedWeb of Science
    15. ↵
      Department of Health. Committee on the Medical Effects of Air Pollution. Non-Biological Particles and Health. London: HMSO, 1995.
    16. ↵
      1. Anonymous
      . Hellenic Ministry for the Environment, Physical Planning & Public Works. http://www.ypeka.gr (accessed 28 May 2010).
    17. ↵
      1. Katsouyanni K,
      2. Touloumi G,
      3. Samoli E,
      4. et al
      . Confounding and effect modification in the short-term effects of ambient particles on total mortality: results from 29 European cities within the APHEA2 project. Epidemiology 2001;12:521–31.
      OpenUrlCrossRefPubMedWeb of Science
    18. ↵
      Institute of Environmental Research and Sustainable Development. Climatological Bulletin. Athens, Greece: National Observatory of Athens (annual), 1998–2004.
    19. ↵
      1. Hastie TJ,
      2. Tibshirani RJ
      . Generalized Additive Models. New York, NY: Chapman and Hall, 1990.
    20. ↵
      1. Wood SN
      . Modelling and smoothing parameter estimation with multiple quadratic penalties. Journal of the Royal Statistical Society, Series B (Statistical Methodology) 2000;62:413–28.
      OpenUrlCrossRefWeb of Science
    21. ↵
      1. Kalkstein LS,
      2. Valimont KM
      . An evaluation of summer discomfort in the United States using a relative climatological index. Bull Am Meteorol Soc 1986;67:842–8.
      OpenUrlCrossRef
    22. ↵
      1. D'Ippoliti D,
      2. Michelozzi P,
      3. Marino C,
      4. et al
      . The impact of heat waves on mortality in 9 European cities: results from the EuroHEAT project. Environ Health 2010;9:37.
      OpenUrlCrossRefPubMed
    23. ↵
      1. Touloumi G,
      2. Samoli E,
      3. Pipikou M,
      4. et al
      . Seasonal confounding in air pollution and health time-series studies: effect on air pollution effect estimates. Stat Med 2006;25:4164–78.
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      R Development Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing, 2005.
    25. ↵
      1. Smith MA,
      2. Jalaludin B,
      3. Byles JE,
      4. et al
      . Asthma presentations to emergency departments in western Sydney during the January 1994 bushfires. Int J Epidemiol 1996;25:1227–36.
      OpenUrlAbstract/FREE Full Text
    26. ↵
      1. Trichopoulos D,
      2. Katsouyanni K,
      3. Zavitsanos X,
      4. et al
      . Psychological stress and fatal heart attack: the Athens (1981) earthquake natural experiment. Lancet 1983;1:441–4.
      OpenUrlCrossRefPubMedWeb of Science
    27. ↵
      1. Katsouyanni K,
      2. Kogevinas M,
      3. Trichopoulos D
      . Earthquake-related stress and cardiac mortality. Int J Epidemiol 1986;15:326–30.
      OpenUrlAbstract/FREE Full Text
    28. ↵
      1. Leor J,
      2. Poole WK,
      3. Kloner RA
      . Sudden cardiac death triggered by an earthquake. N Engl J Med 1996;334:413–19.
      OpenUrlCrossRefPubMedWeb of Science
    29. ↵
      1. Meisel SR,
      2. Kutz I,
      3. Dayan KI,
      4. et al
      . Effect of Iraqi missile war on incidence of acute myocardial infarction and sudden death in Israeli civilians. Lancet 1991;338:660–1.
      OpenUrlCrossRefPubMedWeb of Science
    30. ↵
      1. Wilbert-Lampen U,
      2. Leistner D,
      3. Greven S,
      4. et al
      . Cardiovascular events during World Cup soccer. N Engl J Med 2008;358:475–83.
      OpenUrlCrossRefPubMedWeb of Science

    Footnotes

    • Competing interests None.

    • Provenance and peer review Not commissioned; externally peer reviewed.