Transcript 6A Introduction to airpollution modelling

Modelling of air pollution
-Why?
Magnuz Engardt
January 2008
Swedish Meteorological and Hydrological Institute
Instruments in air pollution assessments
Air quality / deposition measurement programmes
Emission inventories
Effect studies
January 2008
→ Atmospheric transport and dispersion models
Measurements and Modelling
Measure or calculate concentrations and depositions ?
•Models and measurements both have uncertainties
•Some features are particular to either method
•Models and measurements should be used together to explore
January 2008
their full potential -and to increase the quality of each other
Why modelling?
•Mapping of remote regions (incl. areas without measurements)
•Source-Receptor calculations
•Environmental assessments (incl. future / history)
•Find location / consequences of emitters, receptors
•Combine with effect studies (health, acidification, crop yield, …)
•Understand processes in the atmosphere
January 2008
•Check emission inventories
•Verify measurements
•Etc…
January 2008
Some examples…
January 2008
~1000 km
Origin of total non-seasalt sulphur deposition in Sweden
during 1998 as deduced by the MATCH-model
Source-receptor calculations for Southeast Asia
National emissions (Q) and depositions (D) in nine Southeast Asian countries during 2000
500
Q
450
Q
400
D
Gg sulphur per year
350
D
300
250
200
Q
150
D
D
D
100
Q
Q
D
Q
Q
Q
D
Q
D
D
Receptor country
na
m
Vi
et
nd
ai
la
Th
or
e
ap
Si
ng
nm
ar
M
ya
sia
ay
M
al
os
La
ia
es
on
In
d
C
am
bo
di
a
ei
0
Br
un
January 2008
50
Boundary
Shipping
China
Vietnam
Thailand
Singapore
Myanmar
Malaysia
Laos
Indonesia
Cambodia
Brunei
January 2008
Annual total deposition of oxidised nitrogen in South
Asia resulting from NOX emissions in Bangladesh.
~1000 km
January 2008
Climate induced change in total-SOX deposition
(total-, wet- drydeposition) 2021-2050 minus 1961-1990.
Average summer near-surface ozone concentration in
southern Sweden under different emission scenarios
Decrease VOC
or/and NOX emissions
with ~50%
~100 km
January 2008
Other studies include different
NO/NO2 ratio of NOX-emissions
or different speciation of VOC
emissions.
Global distribution of methane
-why does it look like this?
January 2008
Weekly measurements of
“marine boundary layer”
CH4. Data processed by
an interpolating and
“smoothing” program.
January 2008
What is a model ?
Mathematical relations based on empirical or physical laws
Models are used everywhere in society
In our field we have, for example,
Economical models
Numerical weather forecast models
Population models
Climate models
Technological models
Emission inventories
…
Integrated Assessment Models
Dispersion models including emissions, transport,
deposition, chemical conversion etc.
January 2008
...
Estimated change in the global population.
Quality of model output
never better than the input to the model
Meteorology
Atmospheric
Concentration
Emission
Inventory
January 2008
Surface
Deposition
Various
Parameters
Input needed by dispersion models
►Emission data
• Magnitude (and speciation…)
how much is emitted?
• Location (latitude, longitude and height) where is it emitted?
• Temporal variation
how do the emissions vary with time?
►Weather data
• Simple wind-mast
or
• Time varying three-dimensional fields
(historical weather, weather forecasts, weather from climate models, etc.)
January 2008
►Surface characteristics
►Various assumptions
►Etc.
Errors in model results typical due to:
•Emissions wrong
•Meteorology wrong (or too simplified)
•Important processes or parameters are wrong, or omitted,
in the model
•“Bugs” (errors) in the model-code or processing of input/output
(including scaling errors)
January 2008
•Etc.
How good is a model?
Model results must be evaluated in order to assess the accuracy of the model results
 Most common is to compare modelled values with observations
 Mismatch between calculated and observed values can be due to:
• Errors in the model
Note the difference
• Errors in the input to the model
• Errors in measurements
Note the difference
• Non-representative measurements
January 2008
• Etc., …
Model verification
“Objective” statistics, using other measures than mean
and standard deviation often used
Mean error (Bias),
MBE 
RMS-error,
Correlation
r
Ci = simulated value
Mi = measured value
January 2008
N = Number of data points
X = standard deviation of X
X = average of X
1N
 C - Mi
N i1 i
A measure of over- or under estimation.
1 N
2
 (Ci - Mi)
N i1
Gives the magnitude of the error.
1 N (Mi - M)(Ci - C)

N i1
CM
A measure on how well the results co-vary.
Different “objective” measures may give
different scores for a model (!)
Identical meanvalues, no bias
Poor correlation (r≈0)
Large RMS-error
Very different standard deviations.
January 2008
Identical mean-values, no bias
Identical standard deviations
Very poor correlation (r=-1)
Very large RMS-error
Identical standard deviations
Reasonable correlation 0 < r < 1)
Different mean values, high bias
Large RMS-error
Model verification (cont’d)
Visual inspection of results
“Subjective” inspection of the results by plotting them
should also be performed. Methods include:
• Timeseries
• Scatterplots
January 2008
• Maps
Visual inspection of model results
Timeseries
Ozone daily max concentration at Diabla Gora (Polen)
90
Observed daily max
Model daily max
80
70
c(O3) / ppb(v)
60
50
40
30
20
January 2008
10
0
00-02-01
00-03-01
00-04-01
00-05-01
00-06-01
00-07-01
00-08-01
00-09-01
00-10-01
00-11-01
00-12-01
Visual inspection cont’d
25
Scatterplots
2
-3 MATCH
NONO
2 [gm
(µg/m3)] MATCH
All
20
1:1
Skåne
Allerum
Arkelstorp
Klintaskogen
Tunby
15
10
5
0
0
5
10
15
3) Mätdata
NO2 (µg/m-3
20
25
January 2008
NO2 [gm ] Observations
Comparison between calculated and observed monthly average concentrations of
NO2 (g/m3) at four regional background stations. Correlation coefficient R=0,96.
Review of precipitation-chemistry data in India
Data from ~100 stations overlaid MATCH results
January 2008
Underlined digits are suburban stations, others are rural.
Red digits are wet-only collectors, black digits are bulk collectors.
ammonium [μEq l-1]
sulphate [μEq l-1]
Can you use a model of limited quality?
(How “bad” performance is acceptable?)
Unrealistic data should never be accepted
A “factor of two” is often regarded as a very good correspondence
If there is little measured data available you may have to trust your
model results even if the discrepancy is relatively large.
Sometimes you are concerned with typical average levels, sometimes
you want to capture diurnal or day-to-day or seasonal variations
January 2008
Note the problem of unrepresentative measurements
Keep uncertainty in input data in mind (model results could not be
better than the input)
Model quality (cont’d)
It’s good to check the model in different ways
• Both atmospheric concentrations and surface depositions
• Study vertical profiles (although you very seldom have any
data away from the surface…)
• Test both inert and reactive species…
• Both primary and secondary species
• Test the same model at different places and during different
January 2008
periods
If you have discrepancies, try to understand what they are
caused by!
Error propagation
Sometimes small errors in the input cause large errors in
the output
Sometimes it turns out that certain input data or model
formulations doesn’t matter much
Analyse the robustness of your results through sensitivity
January 2008
tests
Atmospheric dispersion modelling
–basic concepts (Ch. 23 in Seinfeld and Pandis, 1998)
Magnuz Engardt
January 2008
Swedish Meteorological and Hydrological Institute
Pollutants (gases and particles) are transported
with the three-dimensional wind
t=t0
January 2008
t=t0+Dt
Note that mean wind and turbulence is not constant
in time or space ! (not even in the tropics)
January 2008
Near-surface wind, pressure and temperature over Sweden 12-24 UTC during 10 September 2007
“Turbulence” cause pollutants to mix and “dilute”
in the atmosphere (Cf. the widening of the plume).
Turbulence is stochastic wind elements (“eddies”)
There are a number of reasons for
turbulence to occur:
● atmospheric (in-) stability
● surface roughness
● vertical wind change
● etc., …
January 2008
The turbulence is varying over time and space.
January 2008
Atmospheric “stability” and surface characteristics
(“roughness” etc.) affects the turbulence
Here the shape of a “plume” during different stabilities
(vertical temperature variations) is illustrated.
Turbulence (and molecular diffusion) may also
transport species in the absence of mean wind
”Closed Chamber experiment” – Molecular
diffusion cause gases to mix.
There is typically no mean vertical wind
close to the ground, still does vertical
transport to and from the surface occur.
This is caused by turbulence.
. .
. .
. .
. .
. .. .. . .
.
.
.
. . . . .. . . . . .. .. . . . . ..
. . .. .
.. ...
. . . .. . . .. .. . . . .. . . . .
.
. . . . .. ..
. ..
. .
.
. . ..
January 2008
CO2 and other gases (O3 ,SO2 …) are taken up by
vegetation. The transport through the stomata of
the leaves occur through molecular diffusion.
Mixed layer, boundary layer
Height
Tracer
profile
Windspeed
profile
Temperature profile
The boundary layer is the part of the
atmosphere that is influenced by
surface friction. Here the atmosphere
is neutrally stratified and tracers are
well mixed. The wind-speed increases
with height; wind-direction also
change with height.
January 2008
Mixed layer
or
Boundary Layer Height.
Typically ~1-2 km during day,
100m or less during night.
Mixed layer height vary over time and space
The depth of the mixed layer height greatly affects nearsurface concentrations
Height
January 2008
Tracer
profile
Windspeed
profile
Temperature profile
A more shallow mixed layer
cause near-surface tracer
concentrations to be higher
Fumigation (downwash)
January 2008
-caused by horizontal variations in near-surface turbulence
(variations in surface roughness and atmospheric stability)
Mixed layer height and temperature profile can
be different over different surfaces due to
different head capacities (land/water) and/or due
to different ”roughness” of the surface.
Local environmental and meteorological effects may
interact with the dispersion of pollutants
January 2008
The spread of a plume during very calm conditions
Even in a flat environment is the wind direction (and magnitude) changing with height
Changing wind direction -and speed- cause
“plumes” not to be straight
~1000 km
100 km
January 2008
Calculated plume of NO2 emitted in Tallinn, Estonia
Dust from Sahara follows trade winds across the Atlantic
January 2008
Different species have different lifetime in the
atmosphere
Species
Lifetime
(Effect in the atmosphere)
“radicals” (OH, H2O2, …)
seconds
Oxidants
Large particles
minutes-hours
(Health,) staining of materials
PM10
a few hours
Health
PM2.5
a few days
Health, Climate
NH3
2-3 days
Acidification, Eutrophication
VOCs
hours-days-weeks-… Health, Near surface ozone
SO2, NOX, O3, …
3-5 days
Acidification, Climate, Crops
CH4, CO
a few months
Climate, near surface ozone
CO2
several years
Climate
CFCs
several decades
Climate, stratospheric ozone
Gases and particles may leave the atmosphere
through drydeposition on various surfaces…
Drydeposition flux is often modelled as:
Fdrydep = vd(z) c(z) [ms-1×gm-3 = gm-2s-1]
January 2008
vd(z) is the ”drydeposition velocity”
and c(z) the concentration of a species at
z meters above surface.
vd(z) is dependent on surface type,
atmospheric stability and is species
dependent.
Dry deposition can be measured through
various more or less advanced methods.
Not routinely done.
Most simple methods include
“throughfall measurements.
Dry deposition can be estimated
through measuring concentrations in
the air and multiplying with relevant
deposition velocities.
Typical drydeposition velocities (valid at 1 m)
Uncertain to at least a factor of two.
Species
Surface type
Time of day
SO2
Grass
Grass
Forest
Forest
Snow
Water
Grass
Grass
Water
Day
Night
Day
Night
O3
NO
NO2
January 2008
NH3
HNO3
Sea
Land
Sea
Land
Day
Night
Day
Night
Value
(cm s-1)
1.2
0.3
0.5
0.2
0.1
1.0
1.0
0.1
0.01
Net source
0.1
0.5
0.1
1.0
5
2-20
January 2008
Drydeposition of particles is a strong function of
particle size
Pollutants can be incorporated in clouds and eventually
be deposited to the ground by precipitation
Scavenging of particles and gases by rain and clouds takes place
during cloud formation, inside clouds and under precipitating
clouds.
January 2008
Scavenging of particles and gases depends on solubility and
cloud and rain type.
Wetdeposition can readily be
measured through collecting
and analysing rainwater.
Species may undergo chemical or physical
transformation
QNO
QNH
X
NQNOX
(1-N)QNOX
(1-S)QSOX
kB•NO3-
NO3-
NO2
JNO2•NO2
k12•O3•NO2
NH3
HNO3
kA•HNO3
D
D,W
SQSOX
kT•fCC SO2
SO42-
Kp=HNO3•NH3
min( NH3 , SO42-)
1.5
Kp=f(RH, T)
irreversible
reversible
NH4NO3
January 2008
X
k21•OH•NO2
k11•O3•NO
NO
QSO
3
D,W
D,W
SO2
kgas•OH • SO2
H0.5(NH4)1.5SO4
D,W
D,W
D,W
D,W
Coupled nitrogen/sulphur chemistry in MATCH
Most reactions depends on ambient conditions (temperature,
abundance of oxidants, solar radiation, humidity etc.).
Physical transformation:
•Gas to particle conversion (or vice versa)
•Particle-to-particle coagulation
•Water condensing on existing particles
January 2008
•Etc.
Summary:
Terms needed during modelling of pollutants:
CONCENTRATION
=
CHANGE
= EMIS + ADVXY + ADVZ + CONVZ + TURBZ + CHEM +PHYS + DRYDEP + WETDEP
EMIS = Emission; release of pollutants into the atmosphere
ADV = Advection; transport with mean wind
CONV = Convective transport; “subgrid” vertical transport in convective clouds
TURB = Turbulent transport; “subgrid” vertical (near-surface) transport due to turbulence
CHEM = Chemical formation/destruction
PHYS = Physical formation/destruction
January 2008
DRYDEP = Drydeposition of gases or particles
WETDEP = Wetdeposition of gases or particles
January 2008
An example from real life:
The Chernobyl accident 25 April 1986
Trajectory calculations
depicting the path of the first
emitted cloud of radioactive
particles from the exploded
Chernobyl reactor.
January 2008
Note that different levels of the
cloud travelled different
routes.
January 2008
Chernobyl accident (cont’d)
Chernobyl accident (cont’d)
January 2008
Measured deposition of 137Cs and rain amount in Sweden
January 2008
Different types of models…
Box-model
Concentrat ion proportion al to
Emissions
Mixed layer height  Horizontal wind speed
It’s possible to create air-pollution indexes
or
Calculate average concentration in a city if
the area and total emissions are known
January 2008
Boundary layer height
Gaussian model
(assume “normal distribution” of pollutants on average)
Instantanoues extent of the
plume at different times
January 2008
When averaging over time
the plume is approximately
normally distributed in the
horizontal and vertical
along the “centre line”
Gaussian model
The Gaussian Plume model (no uptake at the ground at the ground):
2 

2
2 

 1  y     1  z - H  


1
z

H


  exp - 
   exp  - 
 
c( x, y, z) 
exp  - 


 
u2z y
 2   z  
 2   y     2  z  




Q
where Q is the source strength, H is the effective plume height, u the
effective transport velocity,z and y are the vertical and horizontal
dispersion parameters, z the height, y the crosswind distance.
January 2008
z and y are function of stability and distance from the source
Statistical Gaussian models
● Calculate the dispersion from a number of Gaussian plumes.
● Run the model for a number of wind- speeds and directions.
● Add all plumes together.
● The turbulent mixing comes
from z and y. They can be
estimated from wind-profile
January 2008
data and surface characteristics
CFD (Computational Fluid Dynamics)
Cross-section of the plume.
January 2008
A plume from a stack.
Near surface concentrations of pollutants in different industrial areas.
Lagrangian models
Consider an air-parcel that is travelling with the time-varying
three-dimensional wind.
January 2008
Time varying threedimensional wind field
Lagrangian models (cont’d)
January 2008
c ( x, y, z ) 
m
 
2  xy 2 Dz






  xy 

exp - 0.5
x
2



Puff model
Simulate ”dilution” (turbulent
mixing) through making the airparcel
larger.
E.g.: Double the volume will half the
concentration.
Particle model
Simulate ”dilution” (turbulent mixing) through
follow a number of ”particles” which are
spread randomly according to stability etc.
Each “particle” carries a certain mass (which
decreases every time new “particles” are
emitted). After a number of timesteps it is
possible to “add up” the particles in a certain
volume to get the concentration.
Lagrangian models (cont’d)
January 2008
Typical regional spread
from an instantaneous
point-source located
near the surface
Lagrangian models (cont’d)
Lagrangian models
(SO 2 )new  (SO 2 )old  EMIS SO2 - DEPSO2 - CHEM SO2
May include emissions, deposition
(SO 4)new  (SO 4)old  EMIS SO4 - DEPSO4  CHEM SO2
and simple chemistry. More
difficult, however, to include
chemistry where several simulated
species interact.
Lagrangian models are relatively
fast on a computer. Need access to
January 2008
meteorological data.
January 2008
Eulerian models (or gridpoint models)
Eulerian models
Eulerian models divide the
atmosphere into a number of
“gridboxes” and treat advective
and turbulent transport between
boxes, chemistry between species,
emission depositions etc.
The driving data (emissions
meteorology, boundary conditions
etc. varies in time and space.
January 2008
Eulerian models are relatively timeconsuming on computers.
Eulerian models (cont’d)
Eulerian model can cover small areas (cities), regions,
countries, and even the whole globe.
January 2008
The resolution is the “size of the gridboxes”
Eulerian models (cont’d)
Not straightforward to construct advection and chemistry
schemes that are shape and mass conservative etc.
A number of processes, that can not be explicitly
January 2008
described needs to be “parameterised”
Horizontal scale of various air pollution models
January 2008
Model type
Gaussian
CFD
Lagrangian
Eulerian
Macroscale
Mesoscale
Microscale
Global
Regional-to-cont. Local-to-reg.
Local
x
x
x
x
x
x
x
x
x
x
Horizontal scale of various air pollution problems
Scale of dispersion phenomenon
Environmental issue
Local-to-reg. Local
Climate change
x
X
Ozone depletion
x
x
Tropospheric ozone
(x)
x
x
Acidification
(x)
x
(x)
x
x
x
Urban air quality
x
x
Industrial emissions
x
x
Corrosion
January 2008
Global Regional-to-cont.
Basic meteorology…
Chapter 1. in Atmospheric Chemistry and Physics
(Seinfeld and Pandis, 1998)
January 2008
Magnuz Engardt
Do you know…?
What the atmosphere is?
Why the is wind blowing?
Why does it rain?
Why is it colder at night than during day
Why do different regions have different climate?
Why is the sky blue?
How can it be possible to calculate what the weather will be like tomorrow?
Why are the forecasts not always right?
January 2008
What does meteorology has to do with air quality and air pollution?
The atmosphere consists of a mixture of gases
and particles (liquid and solid)
January 2008
The main constituents of the “dry” atmosphere (volume %)
Nitrogen
N2
78.1%
Oxygen
O2
20.9%
Argon
Ar
0.93%
Carbon dioxide
CO2
~0.04% [380 ppm(v)]
Neon
Ne
0.0018%
Helium
He
0.00052%
Methane
CH4
~0.00018% [1.8 ppm(v)]
Krypton
Kr
0.00011%
…
…
…
Near-surface Ozone
O3
~0.000005% [50 ppb(v)]
Sulphur dioxide
SO2
<0.0000001% [1 ppb(v)]
…
…
…
The atmosphere also contains 0-30 g H2O vapour m-3 (0-3%) and 0-1 g H2O particles m-3 (0-0.1%)
The atmosphere
divided into “spheres” depending on the temperature
variation with height.
The pressure is “the weight” of
the air above a certain level.
pV  nRT  p  nRT
V
The pressure at a certain level is proportional
to the number of molecules per volume of air.
 99% of the atmosphere resides under 30 km.
January 2008
Virtually all “weather” (clouds, rain, monsoon
circulation, tropical and extratropical
cyclones, etc.) occur in the troposphere.
Long-lived gases (N2, O2, Ar, (CFCs, N2O, CO2,
CH4),…) are well mixed up to ca. 100 km.
January 2008
The Earth radiation balance
The driving force of weather, (ocean currents,)
and climate
Low latitudes receive more solar
energy per area unit than high
latitudes.
The earth has an energy surplus around the
equator and a deficit near the poles.
January 2008
The earth emits (longwave)
radiation relatively uniformly.
General circulation (distributes heat (energy) from
lower latitudes towards the poles)
Warm air rises near the equator,
Colder air is being “sucked in”
ITCZ (the Intertropical Convergence Zone)
follows the sun between the tropical circles
→ rainy seasons
The earth rotation deflects the air’s movement
→ the trade winds
→“West wind belt” at the mid-latitudes.
Mountain chains and land/sea differences also
have an influence on the circulation
January 2008
Rising air generates clouds
Sinking air causes dry-up -> deserts.
Global maps of surface winds and pressure
during different seasons
January
Note the seasonal shift
of the intertropical
convergence zone, ITCZ
January 2008
July
January 2008
Annual average latitudinal distribution of precipitation, r
(solid line) and evaporation, E (dashed line)
Rotation of the earth affects wind-direction
January 2008
The driving force of winds is
pressure differences.
The rotation of the Earth
deflect the air to the right (on
the N. Hemisphere)
The “Coriolis force”
The wind blows roughly
parallel to the “isobars” in
the “free atmosphere”
Where the surface pressure
is low, the air converges and
is forced upwards.
In high pressure systems, air
diverges, this cause sinking
motion, i.e. “subsidence”.
When “surface friction” is apparent (i.e. close
to the ground) the wind has a component cross
the “isobars”
Generation of sea-breeze
(and monsoon circulation)
((and global general circulation))
Morning (/spring)
Warm air has lower density than
cold air
p
Horizontal temperature variations
cause horizontal pressure variations
winds
p-Dp
p+Dp
Early day (/summer)
p-Dp
p
p+Dp
Mid day (/summer)
p
Height
January 2008
p-Dp
p+Dp
Water
Land
Height
The sea-breeze (summer monsoon) circulation
January 2008
Water
Land
Again, the Coriolis force (and mountain chains etc.) will
deflect the wind from its “original” direction from high
pressure to low pressure
January 2008
Local topographical, or physical properties may
influence wind direction and speed.
Obstacles can affect wind direction as well as
enhance or decrease the wind speed
January 2008
Local meteorology and surface characteristics
determine the planetary boundary layer height.
Various sources of information are used to
describe the current state of the atmosphere
January 2008
Weather radar
Synop stations
Weather satellite
Ordinary physical laws can be used to create a threedimensional picture of the state of the atmosphere
F=mg (Newton’s second law)
pV=nRT (ideal gas law)
Radiation laws (I=T4, etc.)
RH=w/wmax
Conservation of mass
January 2008
Etc.
Analysis and Forecast models
Models are used to fill the gaps between the observations
“Analysis”
Models can also be used to calculate the future state of
January 2008
the atmosphere (weather forecasts)
January 2008
Surface analysis
January 2008
The end…