Particles size and composition in Mediterranean countries:
geographical variability and short-term health effects
MED-PARTICLES Project 2011-2013
Under the Grant Agreement EU LIFE+ ENV/IT/327
Particles size and composition in Mediterranean countries:
geographical variability and short-term health effects
MED-PARTICLES
ACTION 12.
Effect modification of Saharan dust advection days and of forest fires days on the association
between particles and health endpoints: city-specific results and meta-analysis
Report on: Effect modification of forest fires days
Summary: Report on the methods and results of the identification of forest fires events and the investigation
of the effect modification in the Mediterranean cities involved in the MED-PARTICLES project.
-------------------------------------------------------
1
Introduction
Exposure to Forest fires has peculiar characteristics: it is rare and short-lasting; it shows extremely high
levels of combustion-related pollutants and temperature; finally, in the Mediterranean area, fire exposure
has a summer seasonality and it often overlaps with Saharan advection days.
Assessing the population exposure to air pollutants during forest fires present further operating
difficulties: the detection of fire events is currently committed to the fire services, which notice
information about dates, location, duration and extension of the burnt areas. Besides, satellites data and
the relative models offer the opportunity of monitoring the fire emissions, assessing both their local and
long-distance spreading and estimating the concentrations of PM and gaseous pollutants at background
during fires. The network of the fixed monitors could not be able to detect forest fires since they are
located in the largest cities and aimed to monitor the levels of anthropogenic pollutants, especially the
traffic related ones; this makes it difficult detecting the contribution that fire episodes or the re-suspended
combusted matter give to air pollutants.
Also the assessment of fire intensity is not a simple task. Fire-workers classified it according to the size of
burnt area, while the approach using satellite data supply estimates of the fire load on PM10, PM2.5,
sulphate or gaseous levels at the background.
The health effects of fires are due to particulate matter (especially the fine PM), but also to combustion
itself, inorganic gasses, volatile organic compounds (VOC) and to temperature increase during fires.
Mortality has been the most studied effect of this exposure together with respiratory symptoms,
respiratory exacerbations and CV diseases.
We studied, in the framework of the Life+ Med-Particles project, the short-term effects of forest fires on
mortality in 13 European cities of Mediterranean area (in Spain, France, Italy and Greece).
Methods
Exposure assessment
Forest fires events were identified by the Institute of Environmental Assessment and Water Research,
IDÆA-CSIC, Barcelona, Spain, by means of data and models from US NAAPs (Naval Research
Laboratory Marine Meteorology Division), for the Southern Mediterranean cities included in the MedParticles project, in the period 2003-2010. For each city located near the forest fire area the single fireaffected days were identify and the intensity for each day has been assessed as low (PM10 contribute
between 8 and 16 µg/m3, medium (PM10 between 16 and 32 µg/m3) or severe (PM10 above 32 µg/m3).
An additional definition of fire intensity took into account the length of forest fires, classifying them in
2
sporadic episodes, lasting one day, intermediate lasting (2-4 consecutive days) and long lasting (5 or more
consecutive days). Finally, fire-affected day were classified according to their overlap with Saharan days.
The validation of the fire data relied on indirect methods, such as the consistent detection of fire episodes
from satellite data in a same period and area than NAAPs did. An additional approach uses the increases
in mean daily temperature, PM10 levels from fixed monitors and gaseous levels related to combustion
such as carbon monoxide (CO).
Health data
We studied natural, cardiovascular and respiratory mortality as effects of fire days or of PM10 increase (10
µg/m3) during fire days, compared with days without fires. Daily counts of deaths were collected for 13
cities in Italy, Spain, Greece and France, from 2001-2010, for 3-8 years.
Analysis
Percentage increased risk (%IR) of natural, cardiovascular and respiratory mortality for residents (all
ages) were obtained for each city from Poisson regression models, using a case-crossover approach. The
short-term effects of a 10µg/m3 increase of PM10 were estimated within the subsequent six days, adjusting
for time trend, temperature and barometric pressure, and Saharan advection days. We tested the presence
of effect modification of forest fires events on the PM-mortality association adding an interaction term
between PM10 and the binary indicator for fire in the model. Pooled estimates were obtained from a
random-meta-analysis, for 10 cities (excluding 3 cities in Emilia-R region, with 3 fire-days in 3 years,
during spring). We calculated then the p-value of the relative effect modification (p-REM) for the pooled
results and the I2 statistic of Higgins.
Results
The fire days numbered 358 (1.8% of studied days); the cities most affected were Thessaloniki (6%),
Athens (4%) and Rome (3%); the fires intensity was medium-severe in 26% of days. (Table 1).
Considering also the length of episodes, up to 38% of fire days were at high intensity (Table 2). Most fire
days occurred in summer in all cities, except in Barcelona, with a clear winter peak and
three cities in the northern Italy, with the most fires in spring (Figure 1). Thirty five percent of
forest fire days occurred during Saharan dust days, and 10% occurred during a Saharan PM10 increase
over 50% at ground level. The longest period of overlapping between fire and Saharan dust was observed
in Madrid and Palermo (59% of days); in the latter city the overlap between fire and Saharan dust over
50% at ground level get the highest value (43% of days) in the hot season (April to September) (Figure
2). Mean daily temperature increases by 1.6 C° during forest fires in both hot (April to September) and
cold (October to March) seasons. Palermo and Athens showed the highest and consistent increases (Table
3
3). PM10 concentrations increased by 7µg/m3 with higher values in cold season to suggest a more
important contribution of other sources such as traffic and heating in the cold season. Marseille and the
three cities of northern Italy did not show a clear difference between fire-affected and not affected days in
both seasons. (Table 4)
Table 1. Distribution of forest fire days by intensity, season, contemporaneity of Sararan days, in 13 European mediterranean cities,
fire days
study
period
Location
studied
days
N
fire intensity
Madrid
2004-2009
Barcelona
2003-2010
Marseille
2001-2008
Turin
2006-2010
Milan
2006-2010
Bologna
2006-2010
Emilia-R (3 cities) 2008-2010
Roma
2005-2010
Thessaloniki
2007-2009
Palermo
2006-2009
Athens
2007-2009
1735
2922
2922
1826
1826
1826
1096
2191
1096
1461
1096
Tot
34
47
36
14
14
14
3
54
64
34
44
Total
Total (after excluding ER)
19997
19997
358
355
% study
days
2.0
1.6
1.2
0.8
0.8
0.8
0.3
2.5
5.8
2.3
4.0
mild
25
42
26
8
8
8
3
40
43
33
30
mediumsevere
9
5
10
6
6
6
0
14
21
1
14
1.8
1.8
266
263
92
92
season
warm
cold
(April- (OctSept) March)
30
4
18
29
28
8
14
0
14
0
14
0
3
0
53
1
53
11
28
6
42
2
297
294
61
61
Saharan days
dust
dust
no adve- increas increas
ction e <50% e >50%
14
15
5
31
10
4
28
2
2
8
4
1
8
4
0
8
4
1
3
0
0
31
17
5
47
13
3
10
5
15
29
14
1
217
214
88
88
mi
0
2
4
1
2
1
0
1
1
4
0
37
37
16
16
Table 2. Fire episodes by lenght and intensity in 10 Med-Particles cities
Athens
days
Tot
N
18
34
28
47
22
36
8
14
8
14
8
14
18
54
18
64
14
34
10
44
Total
152
Madrid
Barcelona
Marseille
Turin
Milan
Bologna
Roma
Thessaloniki
Palermo
355
1-day episode
mild intensity days
N
N
%
N
7
6
86%
7
18
17
94%
18
16
12
75%
16
4
3
75%
4
4
2
50%
4
4
3
75%
4
8
8
100%
8
9
8
89%
9
8
7
88%
8
2
2
100%
2
2-4 day episodes
mild intensity days
N
N
%
N
27
11
6
55%
22
9
6
12
5
4
10
4
3
75%
10
4
3
75%
10
4
3
75%
17
6
5
83%
13
5
4
80%
6
3
3
100%
12
6
5
83%
5+ day episodes
mild intensity days
N
N
%
N.
80
57
68
85%
80
42
74%
139
7
8
1
1
0
0
4
4
3
2
1
0
1
0
25%
29
42
33%
20
15
2
13%
30
136
4
Table 3. Mean daily temperature observed in days without fire and predicted in fire-affected days
in whole year, hot and cold season, adjusted for month, in the Med-Particles cities.
whole year
in hot season
in cold season
no fire-days
mean
T
Madrid
Barcelona
Marseille
Turin
Milan
Bologna
Roma
Thessaloniki
Palermo
Athens
14.9
14.3
15.5
12.6
13.9
14.7
15.6
18.5
18.5
18.5
Total
15.7
sd
7.6
6.3
7.1
8.0
8.3
8.6
6.9
0.2
0.1
0.2
fire-days
mean T
increase
1.5
1.7
1.6
0.6
0.8
1.2
0.7
2.2
2.9
2.6
1.6
95% CI
0.49 - 2.61
0.89 - 2.51
0.56 - 2.59
-0.98 - 2.27
-0.79 - 2.39
-0.54 - 2.90
-0.03 - 1.53
1.42 - 2.89
2.07 - 3.75
1.76 - 3.51
no fire-days
mean
T
21.1
18.9
20.9
19.1
20.8
21.6
21.1
24.1
22.9
24.1
sd
fire-days
mean T
increase
5.3
4.3
4.5
4.5
4.7
5.0
4.7
5.2
4.3
5.2
21.5
1.4
2.7
1.7
0.6
0.8
1.2
0.7
2.4
2.9
2.6
1.7
95% CI
0.20 - 2.59
1.54 - 3.94
0.74 - 2.69
-0.98 - 2.27
-0.79 - 2.39
-0.49 - 2.85
0.004 - 1.48
1.58 - 3.16
1.95 - 3.85
1.71 - 3.46
no fire-days
mean
T
9.37
9.8
10.2
6.0
7.3
7.9
10.5
13.3
14.3
13.4
10.1
sd
4.3
4.3
4.9
fire-days
mean T
increase
95% CI
1.1
1.1
1.1
-0.20 - 5.50
1.0
1.3
3.0
3.6
-4.74 - 6.76
-0.01 - 2.23
-1.32 - 3.52
4.7
5.0
5.3
4.2
4.2
3.7
4.3
-0.47 - 3.01
1.05 - 4.93
-0.37 - 7.53
1.7
Figure 1
5
60
Fires days during Saharan advection and dust
days in hot season (April-September)
50
40
30
20
10
0
no advection
dust increase <50%
dust increase >50%
Fires days during Saharan advection and dust
days in cold season (October-March)
30
25
20
15
10
5
0
no advection
dust increase <50%
dust increase >50%
Figure 2
6
Table 4. Mean PM 10 levels observed in days without fires and estimated in fire-affected days in whole year,
hot and cold season, adjusted for month and mean temperature, in the Med-Particles cities.
whole year
in hot season
in cold season
no fire-days
fire-days
daily
Pm10
PM10 sd mean
95% CI
level
increase
Madrid
Barcelona
Marseille
Turin
Milan
Bologna
Roma
Thessaloniki
Palermo
Athens
34.8
35.6
26.3
48.3
47.0
38.5
34.5
47.9
35.3
38.2
Total
38.6
19.9
14.7
10.4
36.5
34.4
21.6
14.1
19.7
16.7
20.7
10.7
10.8
3.0
4.0
6.4
2.4
5.5
7.0
9.4
7.0
6.6
4.1 - 17.3
6.0 - 15.7
-0.5 - 6.4
-17.0 - 25.0
-8.2 - 21.0
-7.9 - 12.6
1.8 - 9.2
2.2 - 11.7
4.1 - 14.8
1.0 - 13.1
no fire-days
fire-days
daily
Pm10
PM10 sd mean
95% CI
level
increase
32.5
35.7
25.9
26.9
28.4
27.6
31.0
42.1
34.9
38.4
32.3
15.6
14.7
9.5
13.5
13.2
10.1
10.2
11.7
13.2
17.7
10.2
11.6
0.4
2.9
5.3
1.1
4.8
4.5
12.7
3.1
5.7
5.1 - 15.3
4.9 - 18.2
-2.6 - 3.4
-6.0 - 11.9
-0.5 - 11.1
-4.3 - 6.5
2.2 - 6.9
1.3 - 7.7
8.5 - 16.8
2.0 - 12.6
no fire-days
fire-days
daily
Pm10
PM10 sd mean
95% CI
level
increase
36.8
36.9
26.7
68.0
65.4
49.1
37.7
53.2
35.7
38.0
44.8
22.8
19.2
11.3
11.7 -10.3 - 33.7
7.4
0.9 - 1.6
3.3
-4.5 - 11
39.8
38.6
24.4
16.4
23.8
19.5
23.1
18.5 -13.0 - 50.0
15.4 1.7 - 29.0
-9.5 -24.7 - 5.6
6.4 -23.3 - 36.1
7.6
We observed an increase of 1.43% in natural mortality (though not statistically significant) and up to
7.88% increase in cardiovascular mortality during fire-affected days; no increase was observed in
respiratory mortality. (Table 5)
We found that PM10 (per 10 µg/m3) increased natural mortality by 0.48%, and up to 1.13% (lag 0-1 days)
on fire-affected days, but without statistical significance. In contrast, PM10 increased cardio-vascular
mortality by 0.71%, and up to 4.89% during fire-affected days, with a clear effect modification by fire on
the PM10-mortality association. PM10 also greatly increased respiratory mortality from 1.96% to 4.16%
on fire-affected days but without statistical significance or effect modification. It is worth underlying that
heterogeneity across cities was very low for the PM effects during fire-affected days for natural and
respiratory mortality
7
Table 5. Random meta-analytic estimates of forest fire effects on natural and cause-specific mortality (all ages) in 10 Mediterranean cities.
Natural mortality, lag 0-1
%
pp-value I2
value
heter (%)
REM
Cardiovascular mortality, lag 0-5
pp-value I2
%
L
U
value
heter (%)
REM
Respiratory mortality, lag 0-5
U
Fire days
fire-affected days 2.40
fire-affected days* 2.41
fire-affected days** 1.43
-0.60
-1.54
-1.70
5.49
6.52
4.65
17.27
40.38
17.98
28
9
28
7.88
7.85
6.26
1.52 14.64
0.52 15.72
0.34 12.53
40.50
47.72
29.51
9
5
17
-6.22 -13.86 2.11
-6.48 -13.89 1.58
-7.15 -14.77 1.15
4.91
0.00
0.00
40
52
62
0.48
0.17
0.79
40.68
9
0.73
0.26
0.00
58
1.92
0.57
47.66
5
41.95
39.45
34.34
0
8
9
13
PM10
3
PM10 (10 µg/m ) adj
0.48
for forest fire days.
in fire-free days † 0.46
in fire-affected days † 1.31
0.17
0.15
-0.99
0.79
0.77
3.65
1.20
%
L
U
pp-value I2
value
heter (%)
REM
L
3.28
0
0.476
0.71
0.61
4.89
0.24
0.13
1.88
1.18
1.08
7.99
0.00
0.00
0.00
54
68
60
0.006
1.96 0.61 3.33 47.65
1.92 0.58 3.29 46.80
4.16 -2.03 10.75 0.00
3
0.45 0.02 0.89 61.19
0.72 0.17 1.27 14.98
31
1
2.03 0.54 3.55 53.23
PM10 (10 µg/m ) ‡
in fire-free days ‡ 0.43 -0.01 0.86 60.49
1
0.65 0.13 1.17 4.29
40
1.98 0.50 3.50 52.50
in fire-affected days ‡ 1.49 -1.11 4.17 42.41
8 0.433
2.80 -1.57 7.36 0.00
51 0.342
4.59 -1.79 11.39 0.00
* adjusted for PM10. ** Adjusted for PM10 and Saharan dust days in two levels; † model with interaction between PM10 and forest fire days.
0
5
5
80 0.498
2
3
90 0.445
‡ model with interaction between PM10 and forest-fires days, adjusted for Saharan dust days.
8
Limits
Fire days detected by using satellite data, were confirmed only by indirect methods, according which a
misclassification of exposure could have affected the estimates in Marseille, Turin. Milan and Bologna.
We were unable to estimate the PM10 from forest fires. The study periods differ in length and year
intervals across the cities. The few days of simultaneous exposure to Saharan dust and fires prevented our
studying a possible effect modification between them.
Conclusions:

We found a clear short-term effect of PM10 on cardiovascular mortality during fire days in European
Mediterranean cities.

We also found that fire days modified the effects of PM10 on cardiovascular mortality in
Southern Europe, which is in part explained by the Saharan days effect.

Other factors, such as a different composition of the natural sources and increasing temperature
during fires, may have played an important role in causing mortality during fires.
9