Model reduction algorithms in atmospheric pollution transport simulations by Mauricio Santillana In collaboration with: Daniel Jacob, Philippe Le Sager, Michael Brenner, Scott Norris, and Yevgenii Rastigeyev Image source: http://www.gsfc.nasa.gov/gsfc/earth/terra/co. 4 th GEOS-Chem Scientific Meeting. April, icio Santillana- Harvard University Concentration Carbon Monoxide (CO) in 2000 at 15,000 ft.
12
Embed
Model reduction algorithms in atmospheric pollution transport simulations by Mauricio Santillana In collaboration with: Daniel Jacob, Philippe Le Sager,
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Model reduction algorithms in atmospheric pollution transport simulations
by Mauricio Santillana
In collaboration with: Daniel Jacob, Philippe Le Sager, Michael Brenner, Scott Norris, and Yevgenii Rastigeyev
Imag
e so
urce
: h
ttp:
//w
ww
.gsf
c.na
sa.g
ov/g
sfc/
eart
h/te
rra/
co.h
tm
4th GEOS-Chem Scientific Meeting. April, 2009Mauricio Santillana- Harvard University
Conc
entr
ation
Car
bon
Mon
oxid
e (C
O) i
n 20
00 a
t 15,
000
ft.
How does GEOS-Chem work?
http:
//ce
lebr
ating
200y
ears
.noa
a.go
v/br
eakt
hrou
ghs/
clim
ate_
mod
el/
4th GEOS-Chem Scientific Meeting. April, 2009Mauricio Santillana- Harvard University
• In Every grid box in Earth’s Atmosphere (3D), All chemical species are solved…• Focus: Standard tropospheric ozone-NOx-VOC-aerosol simulation (111 species being solved)
However, do we need to solve all equations in all grid boxes?
Model reduction idea: Solve full equations only where most needed… Reasonable idea?
Higherconcentrations NO
4th GEOS-Chem Scientific Meeting. April, 2009Mauricio Santillana- Harvard University
Image credits: R.V. Martin (Dalhousie U.)
2
Observations
Examples: Isoprene and PAN
Boxes where production or loss rate of NO > 10^2 mol/cm^3 sec
P. L
e Sa
ger &
M. S
antil
lana
4th GEOS-Chem Scientific Meeting. April, 2009Mauricio Santillana- Harvard UniversityActiveInactive
July 8, 2004. 00Hrs GMT
Where are things happening? All species
Percentage of species where production or loss rate > 10^2 mol/cm^3 sec
P. L
e Sa
ger &
M. S
antil
lana
4th GEOS-Chem Scientific Meeting. April, 2009Mauricio Santillana- Harvard University
July 8, 2004. 00Hrs GMT
Where are things happening? All speciesIdea:• Divide domain in active/inactive regions (for each chemical species).• Solve for species that are active in each box.• Use a computationally cheaper method to calculate concentration for inactive species.
4th GEOS-Chem Scientific Meeting. April, 2009Mauricio Santillana- Harvard University
Why 10 threshold?
Example: OH Typical concentration = 10 mol / cmLife time = 1 sec.
2
6 3
Negligible!
What do we do in inactive regions?Two Ideas:
2. Spatial exponential-decay extrapolation (In progress)
• Assume spatial exponential decay from essential regions
• Calculate concentration of species in non-essential regions at each time step using a distance function to the chemical boundary layer
• Ignore all mechanisms such as: transport, chemistry, deep convection, in non-essential regions
1. Alternative ODE solver(Preliminary results)
• Calculate concentration of species in inactive regions at each time step using an explicit ODE.
• Decouples equations.• Cheaper than implicit!
Explicit solver (analytic expression)
4th GEOS-Chem Scientific Meeting. April, 2009Mauricio Santillana- Harvard University