Ozone in Remote Areas: Seasonal Cycles and trends
A contribution to Photooxidants, Particles, and Haze across the Arctic and North Atlantic: Transport,
Observations and Model.
Anne Lindskog (co-ordinator of TOR-2)
Swedish Environmental Research Institute (IVL)
P.O. Box 47086, S-402 58 Göteborg Sweden
Seasonal cycles and trends in ozone in Sweden are studied within the frame of
"Tropospheric Ozone and Precursors – TRends, Budgets and Policy (TROTREP)", a
project of the Thematic Programme for Environment and Sustainable Development
within the Fifth Framework Programme. The project is also a part of the subproject
"Tropospheric Ozone Research" in EUROTRAC-2. The main objective of TROTREP is
to evaluate, validate and predict the effectiveness of past and future EU air quality
legislation with respect to ozone and its precursors.
Data quality is crucial when trend calculations are based on observations. One way to
deal with this problem is to compare time series of daily averages of ozone from
neighbouring sites. Data from Esrange in Northern Sweden have been compared with
the corresponding data from two other sites in Northern Scandinavia, one in Norway,
Jergul (in 1997 replaced by Karasjuk) and the other in Finland, Pallas. The result
indicates a good agreement in terms of seasonal cycles and the day to day variation.
However, for some years one of the three sites (which one differs among years) has
values that are systematically different compared to the other two, see Figure 1. For one
period, a malfunctioning monitor may have caused the divergent results, but in most
cases we have not yet been able to explain the discrepancies.
At sites situated close to large source areas, NO2 can act as a temporary sink for ozone.
In this case Ox (O3+NO2) would be a better parameter to study. NO2 amounts to about
12 % of the winter average and about 5 % of the summer average of Ox at Rörvik (year
2000). However, at a remote site like Esrange the concentration of NO2 is insignificant
in relation to ozone (Figure 2).
Trajectory integrated NOx emissions have been used to segregate ozone monitoring
data from a number of European monitoring sites (Solberg, 2001). The integrated NOx
emissions were combined with ozone monitoring data from EMEP (European
Monitoring and Evaluation Programme) for the years 1988-1996. Daytime averages of
the ozone data were allocated to each of 10 percentile classes of integrated NOx
emissions (calculated as 30-day running averages through the year). The lowest
percentile class (< 10-percentile) of integrated NOx emissions is considered to represent
the European background air. A clear relationship was found between the integrated
NOx-emissions and the ozone observations at the Swedish sites (Figure 3), with
increasing amplitude of ozone with increasing NOx-emissions. The estimated
background seasonal cycle is similar for the different sites. In polluted air masses the
differences are more pronounced and indicate latitude dependence. The seasonal cycle
shows an ozone deficit in winter and a surplus in summer compared to the background.
The difference in ozone concentration between the lowest and the highest percentile
class of integrated NOx emissions is a measure of the "controllable" ozone.
80
1995
70
Esrange 1995
Pallas 1995
Pallas-Esrange
60
50
Jergul 1995
Jergul-Esrange
40
30
20
10
0
-10
-20
17-dec
31-dec
02-dec
16-dec
30-dec
19-nov
03-dec
22-okt
05-nov
08-okt
24-sep
10-sep
27-aug
30-jul
13-aug
16-jul
02-jul
18-jun
04-jun
21-maj
23-apr
07-maj
09-apr
26-mar
26-feb
12-mar
29-jan
12-feb
15-jan
01-jan
-30
80.0
Esrange 1996
Pallas 1996
Pallas-Esrange
Jergul 1996
Jergul-Esrange
1996
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
-10.0
-20.0
18-nov
04-nov
21-okt
07-okt
23-sep
09-sep
26-aug
12-aug
29-jul
15-jul
01-jul
17-jun
03-jun
20-maj
22-apr
06-maj
08-apr
25-mar
11-mar
26-feb
12-feb
29-jan
15-jan
01-jan
-30.0
Jergul
Pallas
Esrange
Figure 1.
Daily averages of ozone in ppb and the differences in ozone
concentrations between neighbouring sites.
80
70
5
Esrange, 1996
Ozone, ppb, left
axis
4
60
50
3
40
2
30
20
1
10
15 per. Mov. Avg.
(Ozone, ppb, left
axis)
15 per. Mov. Avg.
(NO2, ppb, right
axis)
0
ja
n
fe
b
m
ar
ap
r
m
aj
ju
n
ju
l
au
g
se
p
ok
t
no
v
de
c
0
NO2, ppb, right
axis
70
60
40
Rörvik,1999
35
Ozone, ppb, left axis
30
50
25
40
NO2, ppb, right axis
20
30
15
20
10
5
0
0
ja
n
fe
b
m
ar
ap
r
m
aj
ju
n
ju
l
au
g
se
p
ok
t
no
v
de
c
10
15 per. Mov. Avg.
(NO2, ppb, right
axis)
15 per. Mov. Avg.
(Ozone, ppb, left
axis)
Figure 2.
Daily averages and 15-day moving averages of ozone and NO2 in ppb at
Esrange (upper panel) and Rörvik.
Figure 3.
The seasonal cycles of ozone for different percentile classes of trajectory
integrated NOx emissions. The estimated background seasonal cycle (10percentile of NOx-emission) is given with the purple line (from Solberg,
2001).
Seasonal cycles have been calculated using all the data and daytime data (10:00 – 17:00
local time) from three monitoring sites in Sweden, Esrange (67o53’N; 21o04’E) 19912000, Vindeln (64o15’N; 19o46’E) 1986-2000, and Rörvik (57o23’N; 11o55’E) 19872000. In addition, sector analysis has been applied on the ozone data (1988-1996), using
the 2D back trajectories calculated by NILU.
2
1
2
1
3
4
3
Figure 4.
Trajectory based sectors for Rörvik (left panel) and Esrange.
A comparison among the different years points to a large interannual variation with a
clear dislocation in time of the spring maximum. The average seasonal cycle at both
Esrange and Vindeln has a maximum in April and a minimum in late summer. At
Rörvik the maximum occurs in May and the minimum in November.
The number of observations in each sector differs considerably among months, seasons
and years, which has an obvious impact on the seasonal cycle. At Esrange the largest
difference is obtained between sector 2 (22.5 o – 135 o), where the spring peak occurs in
March and the minimum in September, and sector 3 (135 o – 232.5 o) with a maximum
in April and a minimum in October (Figure 5). The results from Vindeln (Figure 6) are
similar (Lindskog and Kindbom, 2001).
At Rörvik 4 sectors are used. Sector 3 (112.5 o – 270 o) contains trajectories from
Continental Europe and UK and thus representing polluted air. This is clearly
demonstrated by the trajectory integrated NOx emission data. In this sector a first ozone
peak is obtained in May and a second one in July (Figure 5). The minimum occurs in
November. The lowest concentration of ozone precursors is found in sector 4 (270 o –
337.5 o) with air masses originating from W and NW. In this sector a spring maximum
is obtained in April. If one considers sector 4 as a proxy for background, an ozone
deficit is seen in the “polluted” sector in winter, most likely due to a reaction with NO,
and a surplus in summer as a result of photochemical production in the boundary layer.
This is in consistence with the results obtained by NILU using the trajectory integrated
NOx emissions (Figure 3 and 7).
60
60
Rörvik
Esrange
50
50
40
40
30
30
20
Average,
Sector 3
20
10
Average,
Sector 4
10
Sector 1
Average
Sector 2
Average
Sector 3
Average
Ju
l
Au
g
Se
p
O
ct
N
ov
D
ec
Ju
l
Au
g
Se
p
O
ct
N
ov
D
ec
Ja
n
Fe
b
M
ar
Ap
r
M
ay
Ju
n
Ja
n
Fe
b
M
ar
Ap
r
M
ay
Ju
n
0
0
Figure 5.
Ozone seasonal cycles in two different sectors averaged over 9 years
(1988-1996) at Rörvik (left panel) and in three different sectors averaged
over 6 years (1991-1996) at Esrange (right panel).
Figure 6.
Trajectory based sectors for Vindeln (left panel) and ozone seasonal
cycles in three different sectors averaged over 9 years (right panel).
Rörvik 1995, monthly averages of ozone in ppb
50
40
S1
30
S2
20
S3
S4
10
ec
D
ov
N
ct
O
p
g
Se
Au
Ju
l
Ju
n
r
ay
M
Ap
b
ar
M
Fe
Ja
n
0
Rörvik 1995, integrated NOx
25
20
S1
15
S2
10
S3
S4
5
ec
D
ov
N
ct
O
Au
g
Se
p
Ju
l
M
ar
Ap
r
M
ay
Ju
n
Fe
b
Ja
n
0
Rörvik 1995, number of observations
Figure 7.
S5
S4
S3
S2
Ju
l
Au
g
Se
p
O
ct
N
ov
D
ec
M
ar
Ap
r
M
ay
Ju
n
b
S1
Fe
Ja
n
800
700
600
500
400
300
200
100
0
Ozone seasonal cycles in four different sectors (upper panel), seasonal
cycles of integrated NOx emission in the different sectors (middle panel)
and the number of observations in the different sectors (lower panel).
S5="undefined".
Over the last 10 – 15 years a significant decrease in nitrogen dioxide concentrations
during the winter half-year has taken place in Sweden, both in urban air and in
background air (Figure 8 and 9).
100
200000
90
180000
80
160000
70
140000
60
120000
50
100000
NO2, "national" winter
half-year average, µg/m3
% of traffic with catalysts
40
80000
30
60000
20
40000
y = -0.7695x + 25.529
R2 = 0.8892
10
20000
Totale, calculated NOx
emission, ton
Passenger car,
calculated NOx emission,
ton
Linear (NO2, "national"
winter half-year average,
µg/m3)
Figure 8.
99/00
98/99
97/98
96/97
95/96
94/95
93/94
92/93
91/92
90/91
89/90
88/89
87/88
0
86/87
0
Urban, calculated Nox
emission, ton
Calculated yearly emissions in tonnes of NOx in Sweden from passenger
cars, from the entire road traffic, and in urban areas (right value axis).
"National" winter half-year averages of NO2 in µg/m3 calculated as an
average of 15 urban areas in Sweden (left value axis) and percentage of
traffic using catalyst (left value axis). (From Svanberg and Lindskog,
2000).
7
Rörvik, NO2, ppb(v)
y = -0.1311x + 5.9667
2
R = 0.3778
6
y = -0.1734x + 6.1341
R2 = 0.6107
5
Annual average
Winter average
Summer average
4
Winter average *
y = -0.1077x + 4.3801
Linear (Annual average)
R2 = 0.6521
3
Linear (Winter average)
Linear (Summer average)
2
y = -0.0359x + 2.578
R2 = 0.1466
Linear (Winter average * )
1
*) refers to winter season, OctDec year1 + Jan-Mar year 2
0
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Figure 9.
Yearly NO2 concentrations at Rörvik between 1987 and 2000.
During the same period the ozone level as yearly average seems to have remained the
same, in spite of the measures taken in Europe to reduce the emissions of ozone
precursors.
A decrease in emissions could affect the regional ozone levels in two different ways. In
winter, a reduction of NOx emissions and the subsequent reduced titration by NO may
result in an increase of ozone. In summer, the reduction would lead to a reduced
photochemical production of ozone in the boundary layer, primarily affecting the peak
values.
The Mann-Kendall test was applied to ozone observations from five background
monitoring stations in Sweden for the period 1985-1998/1999. Increasing ozone levels
were found at most of the stations in February and March (Figure 10). For Vindeln the
most pronounced increase is found in April. The linear regression analysis using all data
for the period 1990-1998 indicated an increase of about 1.1 ppb/year (at the 99%
confidence level) at Rörvik for February and March and about 1 ppb/year at Vindeln for
April. The annual change of the April averages was reduced to 0.8 ppb/year when data
from 1999 was included. A decrease of about 0.6 ppb/year was obtained at Norra Kvill
(the most southern of the sites) for September.
1.2
Median, Kendall
Slope, linear regr.
Ozone, ppb
0.8
0.8
0.4
0.4
Norra kvill
sep
Rörvik Vavihill Vindeln
may
may
jun
0.0
1.2
N. Kvill
feb
Rörvik Vindeln Rörvik
feb
feb
mar
N. Kvill
aug
0.0
-0.4
-0.4
1985-1998
Vavihill
feb
Rörvik
mar
Vindeln
apr
Vavihill
dec
1990-1998
-0.8
-0.8
Station_month
Figure 10.
Norra Kvill Rörvik
feb
feb
Station_month
Magnitude of annual change of monthly averages where the change was
at least 95% statistically significant. The median change according to the
Mann-Kendall test, as well as the slope from linear regression analysis is
presented.( From Kindbom and Lindskog, 2001)
Any tendency in observed concentrations of ozone as a result of measures taken to
reduce the emissions of ozone precursors is expected to be traceable at Rörvik, since
this site is quite frequently exposed to polluted air masses from Continental Europe.
However, as indicated in Figure 11, no significant downward trend was observed in the
yearly averages for the period 1987-2000. Instead, a significant (p=0.05) upward trend
of 0.25 ppb/year in background ozone is obtained. This increase in ozone is even more
pronounced (0.36ppb/year, p=0.005) when only daytime data is considered (Figure 12).
In this case also the 50%-ile is increasing, 0.26 ppb/year (p=0.02). The changes differ
with season. The winter half-year average has increased with 0.36 ppb/year (p=0.001)
(Figure 13). If this increase is driven mainly by the general increase in large scale
background levels and/or the reduction in precursor emissions remain uncertain. Sector
analyses performed on seasonal values (1988-1996) indicate a significant (p=0.01)
increase of the winter half-year 10th percentile (0.69 ppb/year) in the polluted sector
(Figure 14). In addition, an increase in the winter average values is found (0.34
ppb/year, p=0.1). These results support the hypothesis that the increase in ozone is due
to a reduced NO titration. In the clean sector the only significant (p=0.05) increase was
obtained for the winter 95th percentile, 0.45 ppb/year. No tendency was found in the
sector analyses performed on monthly bases.
60
y = -0.0896x + 50.46
R2 = 0.0123
Rörvik, ozone, ppb(v)
all data
50
40
95%-ile
10%-ile
30
y = 0.1301x + 28.321
2
Average
Linear (95%-ile)
R = 0.1057
20
Linear (Average)
Linear (10%-ile)
10
y = 0.2473x + 9.4099
R2 = 0.3057
0
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Figure 11.
Annual averages, 95th- and 10th- percentiles calculated on ozone
observations at Rörvik, 1987-2000.
70
y = -0.1103x + 56.335
2
R = 0.0097
Rörvik, ozone, ppb(v)
daytime data
60
50
95%-ile
40
10%-ile
50%-ile
30
y = 0.2589x + 32.058
R2 = 0.3712
Linear (95%-ile)
Linear (50%-ile)
20
Linear (10%-ile)
y = 0.3568x + 12.857
10
R2 = 0.4853
0
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Figure 12.
Annual averages, 95th- and 10th- percentiles calculated on ozone
daytime (10 a.m.-5 p.m., local time) observations at Rörvik, 1987-2000.
50
y = -0.0648x + 41.879
R2 = 0.0094
Rörvik, ozone, ppb(v)
daytime data
45
40
Winter average
35
Summar average
30
Annual average
y = 0.1459x + 32.457
R2 = 0.1081
Linear (Winter
average)
25
Linear (Annual
average)
y = 0.3567x + 23.035
Linear (Summar
average)
R2 = 0.6198
20
15
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Figure 13.
Annual, winter and summer averages at Rörvik, 1987-2000, based on
daytime data.
60
Average Winter, S3
RÖRVIK, O3
50
y = 0.4493x + 37.755
R2 = 0.3993
10%-ile Winter, S3
95%-ile Winter, S3
y = 0.5564x + 35.655
R2 = 0.1222
Average Winter, S4
40
10%-ile Winter, S4
y = 0.1691x + 26.919
R2 = 0.0449
95%-ile Winter, S4
30
Linear (10%-ile Winter, S3)
y = 0.3356x + 19.554
R2 = 0.3451
Linear (Average Winter, S3)
20
Linear (Average Winter, S4)
y = -0.0297x + 12.691
R2 = 0.0006
Linear (10%-ile Winter, S4)
10
Linear (95%-ile Winter, S3)
y = 0.6927x + 3.3
R2 = 0.5991
Linear (95%-ile Winter, S4)
0
1988
Figure 14.
1989
1990
1991
1992
1993
1994
1995
1996
The development of winter half-year ozone concentrations in sector 3
(polluted) and sector 4 (clean) at Rörvik.
For Esrange only 10 years of data are available. No significant trend was obtained in the
Mann-Kendall test (1991-1999). However, a linear regression analysis applied to the
data showed a significant (at the 95% confidence level) increase of 0.4 ppb ozone/year
in February for the period 1991-1999. When the data from year 2000 is added to the
data set, the increase becomes smaller.
The same tendency of increasing ozone levels is obtained for the annual and seasonal
averages (Figure 15). In this case the results most certainly reflect an increasing
hemispheric background.
50
Esrange, ozone, ppb(v)
y = 0.3746x + 30.347
R2 = 0.2414
40
30
Winter averages
Summar averages
y = 0.3265x + 30.312
y = 0.2783x + 30.278
R2 = 0.2445
Annual averages
R2 = 0.187
20
Linear (Summar averages)
Linear (Annual averages)
Linear (Winter averages)
10
0
1991
Figure 15.
1992
1993
1994
1995
1996
1997
1998
1999
2000
Annual, winter and summer averages at Esrange, 1981-2000.
The development at Vindeln is quite different. On seasonal bases, the only significant
trend is observed in summer. For the period 1986-2000 the summer half-year average of
background ozone, represented by the 10th percentile, has decreased with 0.31 ppb/year
(p=0.02), mainly explained by decreasing concentrations in July, 0.46 ppb/year
(p=0.01), see Figure 15. For individual months (January, February and March)
significant increases of 0.3-0.4 ppb/year are obtained, also of the 50th and 95th
percentiles. A small but significant increase of the 95th percentile is also found in
September, 0.26 ppb/year, p=0.01.
Vindeln is the only site for which a significant trend appears in the sector analysis of
monthly averages (Figure 16).
40
Vindeln, ozone, ppb(v)
35
y = 0.3841x + 18.521
2
R = 0.4128
30
y = -0.0325x + 14.093
R2 = 0.0052
10%-ile, July
25
10%-ile, Winter
10%-ile, Summer
20
10%-ile, February
15
Linear (10%-ile, July)
Linear (10%-ile,
Winter)
Linear (10%-ile,
Summer)
Linear (10%-ile,
February)
10
y = -0.3126x + 14.417
R2 = 0.367
5
y = -0.4625x + 15.494
R2 = 0.4361
0
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Figure 16.
10th percentiles calculated on ozone observations at Vindeln, 1986-2000.
60
Vindeln, ozone mixing ratios (ppb) in April
50
40
Sector 3
All data
Daytime
Linear (Sector 3)
30
y = 1.2646x + 31.698
R2 = 0.8716
20
10
1988
Figure 17.
1989
1990
1991
1992
1993
1994
1995
1996
Average concentration of ozone for April at Vindeln. The red curve and
trend line are based on all data in sector 3 (the "polluted" sector). The
black curve is based on all the data and the yellow on daytime data.
(From Kindbom and Lindskog, 2001.)
Conclusions
One serious problem in trend analysis, aside from data quality and site represenativity,
is the pronounced interannual fluctuation in observed concentrations, which can conceal
the "true" trend. The variation, also manifested as alterations of the seasonal cycle,
seems to be random, without any trend over the time studied. The results of the sector
analyses indicate that the phenomenon may be related to the frequency and duration of
transport in a certain sector and the time of the year when this transport takes place.
However, the interannual variation is significant also within an individual sector. Thus,
the dislocation in time of the maximum and the variability in concentration can not be
explained by the sector frequency alone. The evaluation of the seasonal cycles from a
few individual years indicates, as expected, that the weather conditions, e.g. the number
of hours with sunshine, affect the ozone concentrations. However, for most of the
monitoring sites, meteorological observations are available on rare occasions, which
certainly enhance the uncertainty of the observation based trend analysis
A significant decrease in nitrogen dioxide concentrations during the winter half-year has
taken place in Sweden, both in urban air and in background air, over the last 10 – 15
years. In contrast, a significant increase in winter half-year ozone is obtained at Rörvik.
This is most likely the result of a reduced NO titration. The same tendency of increasing
ozone levels is found at Esrange for both seasons. In this case the results most certainly
reflect an increasing hemispheric background. The development for Vindeln is quite
different with a significant decrease in summer ozone. Increasing levels are however
found for individual months and percentiles.
Acknowledgements
This work was funded by the EU FP5 project TROTREP (EVK2-CT-1999-00043) and
by the Swedish STINT programme. The 2D back trajectories were calculated by Sverre
Solberg, NILU, who also provided data from the Norwegian EMEP sites. The Finnish
data was provided by Tuomas Laurila, FMI.
References
Kindbom, K. and Lindskog, A. (2001) Ozone in Remote Areas: Trends, EUROTRAC-2 Symposium2000
Proceedings. EUROTRAC International Scientific Secretariat, Garmisch-Partenkirchen. CD ROM and
WIT Press, in press
Lindskog, A. and Kindbom, K. (2001) Ozone in Remote Areas: Seasonal Cycles, EUROTRAC-2
Symposium2000 Proceedings. EUROTRAC International Scientific Secretariat, GarmischPartenkirchen. CD ROM and WIT Press, In press
Solberg, S. (2001) Trajectory analyses of European ozone monitoring data. The TOR-2 Annual Report
2000, The EUROTRAC 2 ISS, Munich, in press.
Svanberg, P-A. and Lindskog, A. (2000) Air Quality in Sweden summer 1999 och winter 1999/00. B1388, IVL Swedish Environmental Research Institute. (In Swedish)