The El Niño–Southern Oscillation (ENSO) not only affects meteorological fields but also has a large impact on atmospheric composition. Atmospheric composition fields from the Monitoring Atmospheric Composition and Climate (MACC) reanalysis are used to identify the ENSO signal in tropospheric ozone, carbon monoxide, nitrogen oxide and smoke aerosols, concentrating on the months October to December. During El Niño years, all of these fields have increased concentrations over maritime South East Asia in October. The MACC Composition Integrated Forecasting System (C-IFS) model is used to quantify the relative magnitude of dynamically induced and emission-driven changes in the atmospheric composition fields. While changes in tropospheric ozone are a combination of dynamically induced and emission-driven changes, the changes in carbon monoxide, nitrogen oxides and smoke aerosols are almost entirely emission-driven in the MACC model. The ozone changes continue into December, i.e. after the end of the Indonesian fire season while changes in the other fields are confined to the fire season.
The El Niño–Southern Oscillation (ENSO) is the dominant
mode of variability in the tropics (e.g. Allan et al., 1996). It not only
affects meteorological fields, but it also has a large impact on atmospheric
composition too, for example on ozone (O
These atmospheric composition changes were found in observations (Chandra et al., 1998; Ziemke and
Chandra, 1999; Fujiwara et al., 1999; Chandra et al., 2007; Logan et al.,
2008) and confirmed by modelling studies (Hauglustaine et al., 1999; Sudo and
Takashashi, 2001; Chandra et al., 2002; Doherty et al., 2006; Chandra et al.,
2009; Nassar et al., 2009) which also tried to quantify the relative
importance of the dynamically induced and the emission-driven atmospheric
composition changes. The reasons for the TCO
For El Niño events with large fires over Indonesia, such as in 1997 and
2006, the TCO
The changes in CO are mainly emission-driven (Logan et al., 2008; Chandra
et al., 2009; Voulgarakis et al., 2010) and of smaller horizontal scale than
the O
Large Indonesian wildfires can affect the air quality over South East Asia.
Aouizerats et al. (2015) investigated how the transport of biomass burning
emissions from Sumatra affected the air quality in Singapore. They found that
21 % of the PM
As part on the EU FP7-funded Monitoring Atmospheric Composition and Climate
(MACC) project (
This paper is structured in the following way. Section 2 describes the MACC
reanalysis and the ENSO signal seen in MACC O
The MACC data assimilation system provides analyses and forecasts of
atmospheric composition and was used to produce a reanalysis of atmospheric
composition covering the years 2003 to 2012, as described in Inness et
al. (2013). O
The anthropogenic emissions for the reactive gases for the MACC reanalysis
were taken from the MACCity (MACC/CityZEN EU projects) inventory (Granier et
al., 2011) which accounts for projected trends in the emissions. For the
aerosol fields they came from the EDGAR database (Dentener et al., 2006).
Monthly biomass burning emission for the years 2003 to 2008 from the Global
Fire Emissions Database version 3.0 (GFED3.0) inventory (van der Werf et al.,
2010) were scaled to daily resolution using MODIS active fire observations.
From 2009 to 2012 daily biomass burning emissions from MACC's Global Fire
Assimilation System (GFAS), version 1.0 (Kaiser et al., 2012) were used. One
advantage of the MACC reanalysis is that it used daily fire emissions, in
contrast to several other studies that used monthly averages. Biogenic
emissions used in the MACC reanalysis were for 2003. They came from a recent
update (Barkley, 2010) of the Model of Emissions of Gases and Aerosols from
Nature version 2 (MEGAN2; Guenther et al., 2006,
The emissions are injected at the surface and distributed over the boundary layer by the model's convection and vertical diffusion scheme. Despite the distribution being very efficient, this is a limitation of the current system that will be addressed in future versions. Experiments have been carried out with a new version that uses injection heights based on the Plume Rise Model of Paugam et al. (2015). They show a significant impact on BC AOD for single large fires; the impact at a global scale is smaller: BC AOD is increased by around 5 %. Most of the injection heights calculated with the Plume Rise Model lie within the boundary layer and only a small fraction of smoke (often from particularly intense and well-studied fires) is injected directly into the free troposphere. The largest smoke transport from the boundary layer to the free troposphere occurs through larger-scale meteorological processes. The lowering of the boundary layer height, when air is advected from land to sea, and strong updrafts in the frontal system have previously been identified as efficient smoke transport mechanisms. Similarly, Veira et al. (2015) has studied the sensitivity of AOD in a global climate model to different injection height parameterisations and the above-mentioned plume rise model, with the conclusion that a simple parameterisation reproduces the average larger-scale distribution sufficiently well.
The MACC models do not contain halogenated species, which would contribute
a small additional loss term to O
Initial validation results from the MACC reanalysis are shown in Inness et al. (2013) and Morcrette et al. (2011) and a more detailed validation can be
found in the MACC reanalysis validation reports available from
The MACC reanalysis was used to construct monthly composites of O
Figure 1 shows the warm SST anomaly over the central Pacific associated with El Niño conditions and the resulting precipitation changes for October, November and December from ERA Interim. Precipitation is increased over the central Pacific and reduced over the western Pacific, Maritime Continent, northern Australia and part of the Indian Ocean. Figure 2 shows that the increased precipitation over the central Pacific and the reduced precipitation over the Maritime Continent are collocated with increased ascent and increased descent at 500 hPa, respectively. At the same time, specific humidity at 500 hPa shows a positive anomaly in the area of increased ascent and precipitation over the central Pacific and a negative anomaly over the Maritime Continent. Cloud cover shows a similar signal to humidity, with a negative anomaly over the Maritime Continent and a positive anomaly over the central Pacific (not shown).
Left panels: SST anomaly in K calculated from ERA Interim as the
difference of El Niño composite minus La Niña composite for October
Left panels: anomaly of vertical velocity at 500 hPa in
mm s
The increased biomass burning emissions related to the lack of rainfall over Indonesia and northern Australia can be seen in the fire radiative power (FRP) anomalies shown in Fig. 3. Increased fire activity can be seen over Indonesia in October and November, but stops by December after the end of the fire season, while a weaker biomass burning anomaly continues in northern Australia into December. Over Brazil decreased fire activity can be seen in October.
Biomass burning (fire radiative power areal density) anomaly in
mW m
Figure 4 shows the IAV of the biomass burning emissions for CO from GFAS v1.0
for the 10 years covered by the MACC reanalysis for the area around Indonesia
(10
Time series of daily CO emissions in Tg (10
Biomass burning emissions in Tg per month for CO, NO
Figures 5 to 8 illustrate the impact that the dynamical and emission-related
changes have on the atmospheric composition fields, by showing the anomalies
calculated from the MACC reanalysis at 500 hPa for O
O
During October and November, both dynamical and emission-driven effects
contribute, and modelling studies (e.g. Sudo and Takahashi, 2001; Chandra et
al., 2002; Chandra et al., 2009) have shown that emissions and dynamical
changes can contribute equally for El Niño years with strong biomass
burning. The O
Figure 6 shows the CO anomalies at 500 hPa calculated from the MACC
reanalysis. These anomalies are more confined to the areas of the biomass
burning anomalies (see Fig. 3) than the O
The same as Fig. 5 but for CO anomaly at 500 hPa in ppb.
The NO
The same as Fig. 5 but for NO
Figure 8 shows the anomaly of smoke AOD at 550 nm calculated from the MACC reanalysis. The largest positive anomaly is found over Indonesia in October and November, corresponding to increased aerosol concentrations from biomass burning emissions. The negative aerosol anomaly over South America in October is related to the reduced fire activity seen in Fig. 3. By December the anomalies have disappeared. Similar AOD anomaly patterns over Indonesia were seen by Tosca et al. (2010) when comparing El Niño and La Niña years for August to October for the period 2000 and 2006 from the MISR and MODIS data.
The same as Fig. 5 but for smoke AOD (BC
The three-dimensional nature of the MACC reanalysis allows us to look at the
vertical distribution of the anomalies in the troposphere. Figures 9 to 12
show height versus longitude cross-sections of O
Vertical cross section of O
Figure 10 shows that CO anomalies are largest in the lower troposphere but
can extend throughout the troposphere over Indonesia and South America.
There is a clear connection to increased CO emissions over Indonesia and
decreased emissions over South America due to changes in biomass burning. By
December the anomalies have all but gone and show that there is no
dynamically induced anomaly, unlike in O
The same as Fig. 9 but for CO in ppb.
Figure 11 shows cross sections of NO
The same as Fig. 9 but for NO
Figure 12 depicts cross sections of smoke AOD and shows that, as for CO and
NO
The same as Fig. 9 but for smoke aerosol in ppb.
To quantify the relative impact of increased biomass burning emissions and
dynamically induced changes on the atmospheric composition fields during El
Niño conditions, two experiments are run for the years 2005 and 2006: one
with normal and one with climatological GFAS v1.0 fire emissions. 2006 was
an El Niño year, and 2005 is used to represent normal to weak La
Niña conditions. The additional experiments use the most recent version
of the MACC system, the C-IFS model (Flemming et al., 2015; Inness et al.,
2015). This model is different to the one used in the MACC reanalysis
(Inness et al., 2013) because it has chemistry routines included directly in
ECMWF's Integrated Forecasting System (IFS). A basic initial validation of
C-IFS fields can be found in Flemming et al. (2015) and Inness et al. (2015)
and more detailed validation of C-IFS can be found in the validation reports
available from
The chemistry scheme implemented in the C-IFS model version used for these
experiments is an extended, modified version of the Carbon Bond Mechanism 5
(Yarwood et al., 2005) chemical mechanism as originally implemented in the
chemistry transport model TM5 (Tracer Model 5) (Huijnen et al., 2010;
Williams et al., 2013; Huijnen et al., 2014). This is a tropospheric
chemistry scheme with 54 species and 126 reactions. For O
The anthropogenic emissions used in the C-IFS runs come from the MACCity emission database (Granier et al., 2011). Biogenic emissions are taken from the POET database for the year 2000 (Granier et al., 2005; Olivier et al., 2003), with isoprene emissions from MEGAN2.1, again for the year 2000 (Guenther et al., 2006). Biomass burning emissions for the runs are either taken from GFAS v1.0 (Kaiser et al., 2012) or from a GFAS v1.0 climatology. This daily climatology was constructed using the GFAS v1.0 data set from 2000 to 2014 (Kaiser et al., 2012; Remy and Kaiser, 2014). Biomass burning emissions for each day of the year were defined as the average of the emissions of the same day of the year for the 15 years of the data set.
The differences between the GFAS v1.0 and climatological GFAS emissions for
the area 10
Time series of CO biomass burning emissions in Tg averaged over the
region 10
The experiments are started on 1 January 2005 and run until the end of 2006.
The first experiment (BASE) uses daily GFAS v1.0 emissions, while the second
experiment (CLIM) uses the climatological GFAS data set described above. We
look at fields from these experiments for October and December 2005 and 2006
to determine
the overall impact of changes to the atmospheric composition fields due to El Niño-related dynamically and emission-induced changes by comparing BASE for the years 2006 and 2005 (BASE06 minus BASE05), changes of atmospheric composition due to differences in biomass burning emissions under El Niño conditions by comparing BASE and CLIM for 2006 (BASE06 minus CLIM06), the impact of the El Niño-induced dynamical changes on atmospheric composition and O
Figure 14 shows time series of the tropospheric CO, O
The top panels of Fig. 15 show the overall impact of changes to the
tropospheric O
Time series of the tropospheric CO
TCO
The importance of the dynamically driven ozone changes was also highlighted by Lin et al. (2014, 2015). Despite large El Niño enhancements to wildfire activity in equatorial Asia, the model sensitivity experiments in Lin et al. (2014) indicated that wildfire emissions are not the main driver of ENSO-related ozone variability observed at Mauna Loa, Hawaii. The dynamically induced eastward extension and equatorward shift of the subtropical jet stream during El Niño plays a key role on observed interannual variability of springtime lower tropospheric ozone at Mauna Loa. These shifts enhance long-range transport of Asian ozone and CO pollution towards the eastern North Pacific in winter and spring during El Niño. Lin et al. (2015) demonstrated a connection between springtime western US ozone air quality and jet characteristics associated with strong La Niña winters. They showed more frequent late spring deep stratospheric ozone intrusions when the polar jet stream meanders southward over the western United States as occurs following strong La Niña winters.
The TCO
Top panels: specific humidity differences at 500 hPa in percent for
October
Figure 17 shows that the total column CO (TCCO) anomalies over Indonesia are
almost entirely emission-driven, in contrast to the TCO
TCCO differences in percent for October (left) and December (right) from the experiments BASE06–BASE05 (top), BASE06–CLIM06 (middle) and CLIM06–CLIM05 (bottom). Red colours indicate positive values and blue colours negative values.
October O
As for CO, the NO
Figure 18 shows O
In this paper O
Two simulations were carried out with the C-IFS model to quantify to what
extent the ENSO signal seen in the atmospheric composition fields was due to
changes in biomass burning emissions or due to dynamically induced
changes, e.g. related to changes in the vertical transport of O
Comparing simulations with daily GFAS v1.0 emissions for the years 2005 and
2006 and a daily GFAS v1.0 climatology of the period 2000 to 2014 showed
that tropospheric CO was almost doubled in September 2006 relative to
September 2005 due to increased fire emissions; NO
The results from this paper show that the MACC system is able to model changes in atmospheric composition fields found under El Niño and La Niña conditions. After a more thorough validation of the MACC atmospheric fields against observations, it could be interesting to investigate the ocean–atmosphere response to ENSO-induced changes in atmospheric composition in a further study. A first step would be to include the aerosol direct and indirect effects through the cloud microphysics in the radiation scheme of the IFS and to look at the feedback of fire-induced aerosols on climate. We would expect positive feedback, i.e. reduced convection due to increased atmospheric stability, as carbonaceous aerosols usually absorb (and thus re-emit) a significant amount of solar radiation in the mid-troposphere, and increased aerosol concentrations also lead to reduced land and sea surface temperatures. Their presence should therefore act to reduce convection and precipitation over the Maritime Continent. Including the aerosols in the radiation scheme will also affect the chemical fields through changes in the UV radiation and hence photolysis rates. A second step could see the coupling of the chemistry and aerosol fields by including heterogeneous chemistry on aerosols. In a final step, the MACC system could be coupled with ECMWF's ocean model to investigate how the forcing from ENSO-induced changes to atmospheric composition fields feeds back into the ENSO dynamics.
MACC atmospheric composition data are freely available from
MACC-II was funded by the European Commission under the EU Seventh Research Framework Programme, contract no. 283576. MACC-III was funded by the European Commission under Horizon2020 as a Coordination & Support Action, grant agreement no. 633080.Edited by: P. Jöckel