monthly data for the 1990s - School of GeoSciences

Download Report

Transcript monthly data for the 1990s - School of GeoSciences

2020 vision:
Modelling the near future
tropospheric composition
David Stevenson
Institute of Atmospheric and Environmental Science
School of GeoSciences
The University of Edinburgh
Thanks to:
Ruth Doherty (Univ. Edinburgh)
Dick Derwent (rdscientific)
Mike Sanderson, Colin Johnson, Bill Collins (Met Office)
Frank Dentener (JRC Ispra), Markus Amann (IIASA)
Talk Structure
•
•
•
Chemistry-climate model: STOCHEM-UM
Several transient runs: 1990 → 2030
the 1990s
•
• comparisons with ozone-sonde data
the 2020s
•
• a believable model (hence the first bit)
• a computationally efficient model
• future emissions
• climate change
• other things?
What are the results telling us?
– how modellers use observations
– what is needed to predict the future?
STOCHEM
•
•
•
•
•
•
Global Lagrangian 3-D chemistry-climate
model
Meteorology: HadAM3 + prescribed SSTs
GCM grid: 3.75° x 2.5° x 19 levels
CTM: 50,000 air parcels, 1 hour timestep
CTM output: 5° x 5° x 9 levels
Detailed tropospheric chemistry
•
•
Interactive lightning NOx, C5H8 from veg.
~1 year/day on 36 processors (Cray T3E)
− CH4-CO-NOx-hydrocarbons
− detailed oxidant photochemistry
Model experiments
•
•
•
•
•
Several transient runs: 1990 → 2030
Driving meteorology
–
–
–
Fixed SSTs (mean of 1978-1996)
SSTs from a climate change scenario (is92a)
•
shows ~1K surface warming 1990s-2020s
Shorter run with observed SSTs 1990-2002
New IIASA* global emissions scenarios:
–
–
Business as usual (BAU)
Maximum reductions feasible (MRF)
Stratospheric O3 is a fixed climatology
Vegetation (land-use) also a fixed climatology
*IIASA: International Institute for Applied Systems Analysis (Austria)
IIASA Emissions scenarios
Global totals – there are significant regional variations
Courtesy of Markus Amann (IIASA) & Frank Dentener (JRC)
Model experiments
Compare with
1990s obs
BAU, observed SSTs 1990-2002
BAU, fixed SSTs 1990-2030
MRF, fixed SSTs 1990-2030
BAU, is92a SSTs 1990-2030
1990
2030
Hohenpeissenberg Ozone-sonde model vs observations
(monthly data for the 1990s)
Chemical tropopause (O3=150 ppbv)
Ozone-sonde data from Logan et al. (1999 - JGR)
Hohenpeissenberg Ozone-sonde model vs observations
(monthly data for the 1990s)
Chemical tropopause (O3=150 ppbv)
Ozone-sonde data from Logan et al. (1999 - JGR)
Hohenpeissenberg Ozone-sonde model vs observations
(monthly data for the 1990s)
Chemical tropopause (O3=150 ppbv)
Model
> obs
Model
< obs
Ozone-sonde data from Logan et al. (1999 - JGR)
Hohenpeissenberg Ozone-sonde model vs observations
(monthly data for the 1990s)
Chemical tropopause (O3=150 ppbv)
Ozone-sonde data from Logan et al. (1999 - JGR)
±1 std dev in obs
Hohenpeissenberg Ozone-sonde model vs observations
(monthly data for the 1990s)
Chemical tropopause (O3=150 ppbv)
±1 std dev in obs
±1 std dev in model
Ozone-sonde data from Logan et al. (1999 - JGR)
Hohenpeissenberg Ozone-sonde model vs observations
(monthly data for the 1990s)
Chemical tropopause (O3=150 ppbv)
Model overestimates
by >1 std dev
Model underestimates
by >1 std dev
Ozone-sonde data from Logan et al. (1999 - JGR)
Hohenpeissenberg Ozone-sonde model vs observations
(monthly data for the 1990s)
Chemical tropopause (O3=150 ppbv)
Identify where and
when the model is
wrong
Ozone-sonde data from Logan et al. (1999 - JGR)
Model overestimates
by >1 std dev
Model underestimates
by >1 std dev
Ny Alesund (79N, 12E), Spitzbergen
Model O3 too low in lower troposphere
for all seasons except spring
Ozone-sonde data from Logan et al. (1999 - JGR)
Resolute (75N, 95W), Canada
Model O3 too low in boundary layer
in summer - autumn
Ozone-sonde data from Logan et al. (1999 - JGR)
Sapporo (43N, 141E), Japan
Surface O3 generally too high
Mid-troposphere in
summer too low
Ozone-sonde data from Logan et al. (1999 - JGR)
Wallops Island (38N, 76W), Eastern USA
Mid- & upper-tropospheric
O3 too low in summer
Ozone-sonde data from Logan et al. (1999 - JGR)
Ascension (8S, 14W), Mid-Atlantic
OK at surface, but…
Major O3 underestimate in
tropical mid-troposphere –
too much destruction?
or not enough sources?
Ozone-sonde data from Thompson et al. (2003 - JGR)
•
•
•
•
•
•
The model has some skill at simulating tropospheric
ozone, but is far from perfect.
Careful comparisons with other gases (NOx, NOy,
etc.) also needed, but there is much less data.
For climate-chemistry model validation, lengthy
climatologies, including vertical profiles are most
useful.
If you want modellers to uses the data, provide it in
easy-to-use formats (we’re lazy!)
MOZAIC (operational aircraft data) and satellite data
are examples of the sort of datasets needed.
If you trust the model, it may be useful for future
predictions…
Model experiments
Compare with
1990s obs
BAU, observed SSTs 1990-2002
BAU, fixed SSTs 1990-2030
MRF, fixed SSTs 1990-2030
BAU, is92a SSTs 1990-2030
1990
Compare changes between
the 1990s and 2020s
2030
1990s
Decadal mean values
BAU 2020s
+2 to 4 ppbv over
N. Atlantic/Pacific
>+10 ppbv
India
A large fraction is
due to ship NOx
Change in surface O3, BAU 2020s-1990s
BAU
MRF 2020s
Up to -10 ppbv
over continents
Change in surface O3, MRF 2020s-1990s
MRF
BAU
BAU+climate change 2020s
Change in surface O3, BAUcc 2020s-1990s
MRF
BAU
BAU+cc
Look at the difference between these
two to see influence of climate change
ΔO3 from climate change
Warmer
temperatures &
higher humidities
increase O3
destruction
over the oceans
But also a role
from increases
in isoprene
emissions from
vegetation?
Zonal mean O3 & ΔO3 (2020s-1990s)
1990s
BAU ΔO3
MRF ΔO3
BAUcc ΔO3
Zonal mean OH & ΔOH(2020s-1990s)
1990s
BAU ΔOH
MRF ΔOH
BAUcc ΔOH
CH4, CH4 & OH trajectories 1990-2030
Current CH4 trend
looks like MRF –
coincidence?
All scenarios show increasing OH
Conclusions
•
•
•
•
•
Model development and validation is ongoing, &
is guided by observations
Anthropogenic emissions will be the main
determinant of future tropospheric O3
− Ship NOx looks important
Climate change will introduce feedbacks that
modify air quality
We can estimate the radiative forcing implications
of air quality control measures
NB: Many processes still missing