close

Вход

Забыли?

вход по аккаунту

?

asl.111

код для вставкиСкачать
ATMOSPHERIC SCIENCE LETTERS
Atmos. Sci. Let. 6: 164–170 (2005)
Published online 7 October 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/asl.111
Comparison and visualisation of high-resolution transport
modelling with aircraft measurements
F. M. O’Connor,1 *,† G. D. Carver,1 N. H. Savage,1 J. A. Pyle,1 J. Methven,2 S. R. Arnold,3 K. Dewey4 and J. Kent4
1 NCAS-ACMSU, Centre for Atmospheric Science, University of Cambridge, UK
2 Department of Meteorology, University of Reading, UK
3 Institute for Atmospheric Science, School of Earth and Environment, University
4 Meteorological Research Flight, UK Met Office, UK
*Correspondence to:
F. M. O’Connor, Hadley Centre,
Met Office, Fitzroy Road,
Exeter, EX1 3PB, UK.
E-mail:
fiona.oconnor@metoffice.gov.uk
† Now
at the Hadley Centre, Met
Office, Exeter, UK.
Received: 4 April 2005
Revised: 23 July 2005
Accepted: 23 July 2005
of Leeds, UK
Abstract
This article presents a case study of a comparison of an Eulerian chemical transport
model (CTM) and Lagrangian chemical model with measurements taken by aircraft. Highresolution Eulerian integrations produce improved point-by-point comparisons between
model results and measurements compared to low resolution. The Lagrangian model
requires mixing to be introduced in order to model the measurements. Copyright  2005
Royal Meteorological Society
Keywords: Eulerian model; resolution; aircraft measurements, Lagrangian model; pointby-point comparison
1. Introduction
Global three-dimensional chemical transport models
(CTMs) are often compared with observations from
field campaigns. Driven by meteorological analyses,
they potentially offer a better comparison than general
circulation models (GCMs). Comparison with observations from balloon profiles and aircraft measurements is an important means of validating the chemical
schemes used in such models. However, these model
calculations are subject to a number of errors such as
inaccuracies in the advection and convection schemes
and in the case of CTMs, errors in the meteorological analyses themselves. Another source of error is
model resolution. Despite the mounting evidence that
fine scale structure in the atmosphere is ubiquitous
(Newell et al., 1999), CTMs tend to be run at horizontal and vertical resolutions of 200–500 km and
1–2 km respectively due to computational constraints;
coupled chemistry-climate models are usually run at
even coarser resolutions.
Low-resolution global models effectively mix too
quickly wherever they cease to resolve small spatial
scales. As a result, the production of OH radicals
(POH) in the upper troposphere can be overestimated
by 5–20% at resolutions ranging from 200 to 800 km
approximately (Crowther et al., 2002). Crowther et al.
(2002) studied POH in the upper troposphere while
Esler et al. (2004) examined the impact of spatial
averaging on production and loss of ozone (PO3 and
LO3), loss of CO (LCO), and OH radical concentrations using flight measurements and model integrations
carried out at different resolutions. From the flight
Copyright  2005 Royal Meteorological Society
measurements, they found that PO3 was overestimated
by up to 20% at 3◦ resolution owing to the direct effect
of reducing the negative correlation between NO and
HO2 . LO3 was also increased by grid averaging, but
to a lesser extent. In the model simulations, they found
that differences due to grid averaging were evident in
the tropical upper troposphere and in the winter, extratropical lower troposphere. Furthermore, they argued
that a horizontal resolution of 1◦ was required before
errors in PO3 and LCO, introduced by the misrepresentation of mixing on the lower resolution grid, are
less than 10%. Given these studies, there is clearly a
strong need to run tropospheric chemistry models at
horizontal resolutions of 1◦ or better. However, there is
also a need for improvement in the vertical resolution
of global models. Brunner et al. (2003), in an evaluation of CTMs using aircraft measurements, noted
that high vertical resolution was needed to adequately
represent the steep concentration gradients across the
tropopause.
As Eulerian global models up to now have not
been able to resolve the small-scale structure of
the atmosphere, an alternative approach to using
high-resolution gridded models has been the use of
reverse domain filling (RDF) trajectories (Sutton et al.,
1994) and running detailed chemical box models
along them (e.g. Evans et al., 2000). Although this
approach can successfully reproduce high-resolution
aircraft measurements in some cases (e.g. Methven
et al., 2003), it is subject to uncertainty because it
requires some explicit representation of mixing along
trajectories, and the timescale for mixing is not unique
(Good et al., 2003).
Comparison of high-resolution transport modelling with aircraft measurements
There are a number of difficulties in comparing gridded model results with flight data. Large differences in
spatial scales between model and measurements can be
dealt with by aggregating the flight measurements and
comparing in some statistical sense. However, Brunner
et al. (2003) showed that ‘point-by-point’ comparisons
(where the model is sampled at the same time and spatial locations to the aircraft flight) may produce more
meaningful statistics. In regions of sharp gradients or
filamentary structures, the resolution of the model will
be important if such features are to be captured adequately for long-lived species whose distribution is
determined by transport.
In recent years, there have been significant improvements in PC hardware that has allowed advanced
graphics techniques, previously only available with
specialised hardware, to be available with Linux-based
PC desktop machines. In addition, there has been significant developments in open source and freely available visualisation and analysis software, both general
purpose and specifically for the environmental sciences; e.g. IDV (http://my.unidata.ucar.edu/content/sof
tware/IDV/index.html), OpenDX (http://www.opendx
.org/) and Vis5D (http://vis5d.sourceforge.net/). In
addition, developments in de facto data standards such
as netCDF and HDF which are accepted by many
available software packages greatly ease the use of
these programs and assist in data exchange.
In this article, we present a case study of a comparison of both Eulerian gridded and Lagrangian modelling techniques with aircraft measurements. We show
the effect of using higher model resolutions in the
comparison. We will also use 3D visualisation techniques to aid our interpretation of differences between
modelled and measured values.
2. The p-TOMCAT model and ACTO
campaign
The model used in this study is the three-dimensional
CTM, p-TOMCAT, a full description of which can be
found in O’Connor et al. (2004). It uses the secondorder moment advection scheme of Prather (1986)
and includes 50 chemical species and 130 reactions.
Parametrisations for wet and dry deposition, boundary layer mixing and convective processes as well
as a comprehensive emissions inventory are included.
It has recently been parallelised using the Message
Passing Interface (MPI) and now offers much greater
capability in terms of resolution and/or complexity.
Here, the model has been run at a number of different resolutions, both horizontally and vertically, in
an attempt to improve the comparison with aircraft
measurements. No chemistry was included in these
integrations; only the transport component was used.
The case study presented here is a single flight of
the UK Met Office (UKMO) C-130 aircraft during
the Atmospheric Chemistry and Transport of Ozone
(ACTO) campaign in May 2000. This flight was
Copyright  2005 Royal Meteorological Society
165
chosen because it included observations of a stratospheric intrusion, straddled on either side by clean
marine boundary layer air and polluted boundary layer
air. Moreover, this flight has been extensively analysed
and interpreted with RDF trajectories and a Lagrangian
chemical box model by Methven et al. (2003).
3. Results
Figure 1 shows modelled ozone on the 383 hPa surface at 12 UT on 19 May 2000; the black line represents the flight track of the UKMO C-130 aircraft
during the ACTO campaign. Figure 1(a) was produced
by advecting a low-resolution ozone field from pTOMCAT for one day at a horizontal resolution of
2.8◦ with 31 vertical levels from the surface up to
the top boundary at 10 hPa. Figure 1(b), on the other
hand, was produced by advecting the same field but
at higher resolution of 1.1◦ with 44 vertical levels.
These 44 levels are the same vertical levels as in
the ERA40 reanalysis project, but the top boundary
in p-TOMCAT remained at 10 hPa. In both cases, the
low-resolution initial ozone field was simply advected;
no chemistry was included. On this timescale, the
advection of a relatively long-lived species, such as
ozone, is valid. In both cases, a stratospheric intrusion
extending southeastwards from Iceland to Scotland is
evident. However, in the high-resolution simulation
(Figure 1b), the intrusion is narrower in horizontal
extent; this matches more closely the morphology of
the intrusion in the Meteosat water vapour channel
image for 12 : 46 UT on that day (see Figure 1 of
Methven et al., 2003). In addition, higher peak ozone
concentrations are evident within the intrusion. Furthermore, the stratospheric intrusion is separated from
the neighbouring uplifted boundary layer air on both
its western and eastern fringes by sharp gradients in
the high-resolution simulation. Although not shown,
a similar improvement in the synoptic representation
of CO is also found, with the high-resolution integration showing stronger horizontal gradients with low
CO in the stratospheric intrusion and enhanced concentrations in the polluted uplifted boundary layer air
on the eastern side of the stratospheric intrusion. A
vertical cross section of ozone at both resolutions can
be found in Figure 2; it indicates the presence of a
tropopause fold, extending into the troposphere in a
westward direction. At high resolution, the tropopause
fold extends further west and penetrates to lower altitudes than in the low-resolution simulation. Sharper
gradients, both in the vertical and in the horizontal
directions, are evident at higher resolution.
The validity of these model simulations and the
impact of increased resolution can be assessed by comparison with the aircraft observations taken on board
the UKMO C-130 aircraft. Ozone was measured by an
ultraviolet (UV) absorption technique with an accuracy
of ±3% and carbon monoxide was measured using
vacuum UV resonance fluorescence with an accuracy
Atmos. Sci. Let. 6: 164–170 (2005)
166
FM O’Connor et al.
(a)
260
70°N
240
66°N
220
200
180
160
58°N
140
O3/ppbv
LATITUDE
62°N
120
54°N
100
50°N
80
60
46°N
40
20°W
16°W
12°W
8°W
4°W
LONGITUDE
0°E
4°E
8°E
(b)
260
70°N
240
220
66°N
200
180
160
58°N
140
O3/ppbv
LATITUDE
62°N
120
54°N
100
50°N
80
60
46°N
40
20°W
16°W
12°W
8°W
4°W
0°E
4°E
8°E
LONGITUDE
Figure 1. Modelled ozone fields on the 383 hPa surface from the p-TOMCAT model at 12 UT on 19 May 2000 at resolutions
of (a) 2.8◦ with 31 vertical levels, and (b) 1.1◦ at 44 vertical levels. The flight track of the C-130 aircraft relative to the modelled
fields is superimposed as a black line
of ±15% (Gerbig et al., 1999). Figure 3 shows the
measured ozone and carbon monoxide concentrations
along the flight track in comparison with the lowresolution modelled fields (Figure 3a,b). It indicates
that the low-resolution simulation cannot accurately
capture the spatial variation observed along the flight
track. It also suggests that point-by-point comparisons of Eulerian model output with high-resolution
aircraft measurements are not an ideal method to
validate global models run at low resolution. Indeed,
most aircraft-model comparisons have been based on
monthly or seasonal mean data, such as those of
Copyright  2005 Royal Meteorological Society
Law et al. (2000) and Bregman et al. (2001). More
recently, however, Brunner et al. (2003) attempted a
more direct approach by comparing each measured
data point with its temporally and spatially interpolated
model counterpart. Results over North America, however, showed that the models tended to underestimate
CO when the observations were high and vice versa,
suggesting that the coarse resolution of the models
does not preserve strong spatial gradients.
By contrast, the higher-resolution advection shows
a marked improvement in terms of capturing the
observed concentrations both within the stratospheric
Atmos. Sci. Let. 6: 164–170 (2005)
Comparison of high-resolution transport modelling with aircraft measurements
167
(a)
260
240
9.0
220
200
180
160
140
5.0
O3/ppbv
Z (KM)
7.0
120
100
3.0
80
60
1.0
20°W
40
16°W
12°W
8°W
4°W
0°E
LONGITUDE
(b)
260
240
9.0
220
200
180
160
140
5.0
O3/ppbv
Z (KM)
7.0
120
100
3.0
80
60
1.0
20°W
40
16°W
12°W
8°W
4°W
0°E
LONGITUDE
Figure 2. Vertical cross section of ozone along 60◦ N from the p-TOMCAT model at 12 UT on 19 May 2000 at resolutions of
(a) 2.8◦ with 31 vertical levels and (b) 1.1◦ with 44 vertical levels. The flight track of the C-130 aircraft relative to the modelled
ozone fields is superimposed as a black line
intrusion and in the neighbouring air masses (Figure
3c,d). Equally, the transition between the neighbouring
air masses is well reproduced. However, ozone concentrations within the stratospheric intrusion in the latter stages of the flight are overestimated, although this
was also evident in the Lagrangian study by Methven
et al. (2003). They attributed it to mixing between
stratospheric and tropospheric air on the lower, eastern portion of the fold that was not represented
explicitly by the Lagrangian model. Here, although
Copyright  2005 Royal Meteorological Society
there is implicit mixing associated with the advection scheme on the Eulerian grid, it is not sufficient
for the modelled and measured concentrations to be
comparable. This comparison, nevertheless, indicates
that higher resolution in model simulations can capture
the concentration fluctuations observed along aircraft
flight tracks which are associated with sampling air
masses of different origins with distinct chemical
composition. Furthermore, it confirms that higher resolutions are required for reasonable point-by-point
Atmos. Sci. Let. 6: 164–170 (2005)
168
FM O’Connor et al.
(a)
O3/ppbv
300
200
100
0
11
12
13
14
15
16
17
16
17
16
17
time of day/hours UT
CO/ppbv
(b)
200
150
100
50
0
11
12
13
14
15
time of day/hours UT
O3/ppbv
(c)
300
200
100
0
11
12
13
14
15
time of day/hours UT
CO/ppbv
(d)
200
150
100
50
0
11
12
13
14
15
time of day/hours UT
16
17
Figure 3. Comparison between measured (black) and modelled
(red) ozone in (a) and (c) and carbon monoxide in (b) and
(d) along the UKMO C-130 flight track on 19 May 2000. The
model simulation was carried out at a horizontal resolution of
2.8◦ with 31 vertical levels in (a) and (b) while (c) and (d) show
the results from the same simulation at a horizontal resolution
of 1.1◦ with 44 vertical levels. The measurements along the
flight track are reported every 10 s. The model outputs values
along the flight track every 15 min. In one hour, the aircraft
travels approximately 420 km. For comparison, at 60◦ N, the
model grid spacing is approximately 25 km for a grid resolution
of 2.8◦ and 10 km for 1.1◦
comparisons of global chemistry models with aircraft
measurements.
By visualising three-dimensional isosurfaces of
ozone from the low- and high-resolution integrations,
it is easier to appreciate the differences in the representation of the synoptic field which helps put the differences seen in Figure 3 into context. Figure 4 shows
the 85 ppbv ozone isosurface from the low and highresolution runs. The model output was regridded to
0.5◦ for visualisation purposes. The stratospheric intrusion is clearly seen (compare with Figures 1 and 2)
with the higher-resolution isosurface extending further
into the troposphere, narrower and positioned more
westward. Figure 5 provides a rotating view of the two
isosurfaces. Note that the higher-resolution integration
produces a second folded structure east of the first that
is completely missing in the low-resolution integration. By employing 3D visualisation techniques in this
way, it has provided further insight into the structure
of the fold in the ozone field in both simulations.
The high-resolution simulation results in a better
comparison with the aircraft measurements. However,
Copyright  2005 Royal Meteorological Society
Figure 4. Isosurfaces of ozone at 85 ppbv. The yellow isosurface
is from the low-resolution (2.8◦ ) integration while the green
isosurface is from the high-resolution (1.1◦ ) integration. The
green isosurface has been made partially transparent. The
bottom of the 3D cube shows a low-resolution topography of
the British Isles for reference. Both integrations were regridded
onto a 0.5◦ grid for visualisation purposes. This plot was
produced by Vis5D
an important question is the extent to which horizontal
and/or vertical resolution is critical. There are a number of studies that argue that horizontal resolution is
important (Crowther et al., 2002; Esler et al., 2004),
whereas other studies suggest that high vertical resolution is critical (Land et al., 2002; Brunner et al., 2003).
Equally, the horizontal resolution of the forcing fields
between the two simulations already considered was
increased from 2.8◦ × 2.8◦ to 1.1◦ × 1.1◦ ; what impact
does the small-scale structure in the forcing analyses affect the simulations? We carried out a number
of integrations with p-TOMCAT to assess the relative
importance of horizontal and vertical resolutions and
resolution of the input analyses of wind, temperature
and humidity. Results from these runs (not shown)
indicate that increasing the horizontal resolution of
the model alone improves the comparison with flight
data but increasing the resolution of the analyses to
1◦ × 1◦ gives larger improvement in the comparison.
Increasing the horizontal resolution to 0.75◦ but retaining 1◦ × 1◦ analyses produced very little improvement.
Improving the vertical resolution had no impact for
this case, most likely because the timescale for the
vertical transport is much longer than the short integrations carried out here.
The time period used for advection was only one
day, as this was found to be an appropriate timescale
by Methven et al. (2003). Given that ozone and carbon monoxide are fairly long lived in the troposphere,
longer timescales could be considered. Figure 6 shows
Atmos. Sci. Let. 6: 164–170 (2005)
Comparison of high-resolution transport modelling with aircraft measurements
169
(a)
300
200
100
0
11
12
13
14
15
16
17
16
17
time of day/hours UT
(b)
300
200
100
0
11
12
13
14
15
time of day/hours UT
Figure 6. Comparison between measured (black) and modelled
(red) ozone along the C-130 flight track on 19 May 2000.
The modelled concentrations are derived using trajectories of
(a) one day and (b) four days
Figure 5. Movie of the isosurfaces of Figure 4 rotating through
360◦ . The user is encouraged to manually move the movie slider
control to view the structure from the desired angles. Note
the difference in the structure of the intrusions at high and
low resolutions. The higher resolution integration isosurface is
further west, narrower in extent and extends further to the
surface. Also note the significantly different structure to the east
of the main intrusion. The high-resolution integration shows a
second intrusion-like feature which is not represented at low
resolution. The individual frames were generated using Vis5D
and combined to form a movie
modelled ozone along the aircraft flight track when
three-dimensional trajectories are used to advect a
low-resolution ozone field from p-TOMCAT on to
the flight track. In Figure 6(a), a time period of one
day was used, which results in a reasonable comparison between modelled and measured ozone although
ozone, as mentioned previously, is overestimated in
the stratospheric intrusion towards the end of the
flight. In Figure 6(b), the length of the trajectories was
increased from one to four days. Here, the Lagrangian
approach generates too much structure at fine scales,
which is not evident in the observations. This is primarily due to the lack of mixing between air masses
along the trajectories. However, even when mixing is
included, significant tuning is required to the rate of
dilution or the background concentrations, since the air
parcel shows strong sensitivity to both of these (Wild
et al., 1996). By contrast, the Eulerian model does
not suffer from these difficulties since the advection
scheme has implicit mixing. Integrations of differing
lengths therefore produce consistent results.
4. Conclusions
The case study presented in this article highlights the
difficulty of comparing flight track measurements with
output from CTMs. A number of difficulties, including
Copyright  2005 Royal Meteorological Society
disparity between the scales of measurements and
model grid, make quantitative comparisons difficult. In
this article, we present evidence that greatly increasing
the horizontal resolution of the model and the forcing
analyses can make a significant improvement in the
comparison. For the example shown here, this is
largely due to a much improved representation of the
synoptic situation. We have considered other cases
where an improvement arises though not as significant
as presented here. The improvement due to increase in
resolution is likely to be specific to each case.
In cases where the flight passes through several distinct airmasses, such as the case here, which samples
tropospheric and stratospheric air, it can be difficult
to understand the differences seen in modelled and
observed flight track data. We have also used 3D visualisation techniques to illustrate how the aircraft flight
path relative to the stratospheric intrusion represented
by the model accounts for the differences seen. We
believe the use of 3D representations of model output
in cases similar to this are a useful aid to interpreting
flight data for modelling purposes.
Acknowledgements
The authors acknowledge the European Centre for MediumRange Weather Forecasts (ECMWF) for allowing access to
their meteorological products and the British Atmospheric Data
Centre for providing them. We also acknowledge Hannah Barjat
for her involvement in ACTO and data processing.
The Atmospheric Chemistry Modelling Support Unit is
one of the distributed centres of the NERC Centres for
Atmospheric Science (NCAS). F.M. O’Connor and G.D. Carver
acknowledge NCAS for funding, and the ACTO aircraft
campaign was funded by the NERC’s upper troposphere/lower
stratosphere (UTLS) thematic programme.
Atmos. Sci. Let. 6: 164–170 (2005)
170
References
Bregman A, Krol MC, Teyssedre H, Norton WA, Iwi A, Chipperfield M, Pitari G, Sundet JK, Lelieveld J. 2001. Chemistry-transport
model comparison with ozone observations in the midlatitude lowermost stratosphere. Journal of Geophysical Research 106(D15):
17 479–17 496.
Brunner D, Staehelin J, Rogers HL, Köhler MO, Pyle JA, Hauglustaine D, Jourdain L, Berntsen TK, Gauss M, Isaksen ISA, Meijer E,
van Velthoven P, Pitari G, Mancini E, Grewe V, Sausen R. 2003.
An evaluation of the performance of chemistry transport models
by comparison with research aircraft observations. Part 1: concepts
and overall model performance, Atmospheric Chemistry Physics 3:
1609–1631.
Crowther RA, Law KS, Pyle JA, Bekki S, Smit HGJ. 2002. Characterising the effect of large-scale model resolution upon calculated
OH production using MOZAIC data. Geophysical Research Letters
29: 12, 10.1029/2002GL014660.
Esler JG, Roelofs GJ, Kohler MO, O’Connor FM. 2004. A quantitative analysis of grid related systematic errors in oxidising capacity
and ozone production rates in chemistry transport models. Atmospheric Chemistry and Physics 4: 1781–1795.
Evans MJ, Shallcross DE, Law KS, Wild JOF, Simmonds PG, Spain
TG, Berrisford P, Methven J, Lewis AC, McQuaid JB, Pilling MJ,
Bandy BJ, Penkett SA, Pyle JA. 2000. Evaluation of a Lagrangian
box model using field measurements from EASE 1996. Atmospheric
Environment 34: 3843–3863.
Good P, Giannakopoulos C, O’Connor FM, Arnold SR, de Reus M,
Schlager H. 2003. Constraining tropospheric mixing timescales
using airborne observations and numerical models. Atmospheric
Chemistry and Physics 3: 1023–1035.
Gerbig C, Schmitgen S, Kley D, Volz-Thomas A, Dewey K, Haaks D.
1999. An improved fast-response vacuum-UV resonance
Copyright  2005 Royal Meteorological Society
FM O’Connor et al.
fluorescence CO instrument. Journal of Geophysical Research
104(D1): 1699–1704.
Land C, Feichter J, Sausen R. 2002. Impact of vertical resolution on
the transport of passive tracers in the ECHAM4 model. Tellus, Series
B 54(4): 344–360.
Law KS, Plantevin PH, Thouret V, Marenco A, Asman WAH, Lawrence M, Crutzen PJ, Muller JF, Hauglustaine DA, Kanakidou M.
2000. Comparison between global chemistry transport model results
and measurement of ozone and water vapor by airbus in-service
aircraft (MOZAIC) data. Journal of Geophysical Research 105(D1):
1503–1525.
Methven J, Arnold SR, O’Connor FM, Barjat H, Dewey K, Kent J,
Brough N. 2003. Estimating photochemically produced ozone
throughout a domain using flightdata and a Lagrangian model. Journal of Geophysical Research 108(D9): 4271, doi:10.1029/2002JD002955.
Newell RE, Thouret V, Cho JYN, Stoller P, Marenco A, Smit HG.
1999. Ubiquity of quasi-horizontal layers in the troposphere. Nature
398: 316–319.
O’Connor FM, Law KS, Pyle JA, Barjat H, Brough N, Dewey K,
Green T, Kent J, Phillips G. 2004. Tropospheric ozone budget:
regional and global calculations. Atmospheric Chemistry and Physics
Discussions 4: 991–1036.
Prather MJ. 1986. Numerical advection by conservation of secondorder moments. Journal of Geophysical Research 91: 6671–6681.
Sutton RT, Maclean H, Swinbank R, O’Neill A, Taylor FW. 1994.
High-resolution stratospheric tracer fields estimated from satellite
observations using Lagrangian trajectory calculations. Journal of the
Atmospheric Sciences 51: 2995–3005.
Wild O, Law KS, McKenna DS, Bandy BJ, Penkett SA, Pyle JA.
1996. Photochemical trajectory modelling studies of the North
Atlantic Region during August 1993. Journal of Geophysical
Research 101(D22): 29269–29288.
Atmos. Sci. Let. 6: 164–170 (2005)
Документ
Категория
Без категории
Просмотров
5
Размер файла
653 Кб
Теги
asl, 111
1/--страниц
Пожаловаться на содержимое документа