1. ACITES Edinburgh: Metrics of model transport
David Stevenson, Sue Liu
(Fiona O’Connor, Martyn Chipperfield, Mat Evans, Oliver Wild)
(+thanks to: Luke Oman, Franz Conen)
ACITES Networking Meeting, York, 8-9th January 2013

2. How good are global atmospheric models?
Recent ACCMIP (and earlier ACCENT, HTAP) model inter-comparisons had ‘limited’ model evaluation…
E.g. ACCMIP O3, CO etc figs Young, Naik

3. Ozone: Sondes, TES vs models
Correlations and bias numbers provide a metric of overall model performance with respect to simulating tropospheric ozone
Young et al, ACPD

4. Tropospheric ozone column: models vs OMI/MLS
Agreement with OMI/MLS also a measure of model performance
Young et al, ACPD

5. CO @ 500 hPa: Models - MOPITT
Similarly, agreement of CO with MOPITT…
Naik et al, ACPD

6. How good are global atmospheric models?
Models overestimate (+20%) NH ozone, underestimate (-20%) SH ozone
Models underestimate (-20 ppb) NH mid-lat CO, overestimate (+30 ppb) tropical CO
(NB assuming we believe observational datasets…)
Why? Emissions, Chemistry, Transport, … ?
These evaluations test many model processes integrated together
Can we isolate processes?
E.g. just evaluate model transport schemes?
But these sort of evaluations do not isolate single processes, such as transport…

7. How can we test a model’s transport?

8. Schematic section through atmosphere to identify some transport processes (e.g. HTAP 2010)

9. Transport and mixing processes
Many different processes:
Large scale advection/mixing
Horizontal and vertical
Sub-grid scale advection:
Convection
Boundary layer mixing
Transport/mixing across significant ‘transport barriers’
Boundary layer – free troposphere
Tropopause
Inter-hemispheric transport
Between air masses at fronts

10. Current Plans Sue Liu to start on 1st April 2013
Identify relevant observational datasets to assess a variety of model transport/mixing processes, e.g.:
Strat-trop exchange: O3/CO/H2O distributions
Tropopause location: tracers, e.g. e90
Convection: 222-Rn, CO, CH3I, LiNOx
Large scale dynamical variability: O3-ENSO index
Horizontal advection: volcanic ash/SO2, Fukushima/nuclear accidents, tracer experiments
BL/free troposphere exchange: 222-Rn profiles
Test within UKCA/other UK models
Test more widely within other international models

11. An appeal to the community
If you have ideas of datasets/methods for testing model transport that you would like to share, please let us know

12. ENSO-Ozone Longitudinal section at equator Latitudinal sections at
Testing large scale dynamical variability
Oman et al. (2011) Geophys. Res. Lett., 38, L13706, doi: /2011GL047865
Latitudinal sections at
different longitudes
Ozone-ENSO Index – it is the monthly mean O3 column for two broad regions of E and W Pacific, differenced.
Closely correlates with ENSO itself.
Column O3 responds to changes in convection and large-scale dynamics.
If model can capture observed behaviour, indicates that large scale response of O3 to changes in dynamics and convection are correctly represented. It is still an integrated metric, as it is unclear how much of the ozone response is due to transport and how much is due to chemistry. Could run an additional passive tracer within such simulations to separate out effects (but can’t do this in the observations of course).

13. 222-Rn
Franz Conen’s data from Bern/Jungfraujoch BL vs Free troposphere

14. 222-Rn: diurnal cycles

15. 222-Rn: vertical profiles*
Williams et al 2011

16. Black Saturday Fires Australia Feb 2009
Anomalous CO detected
Fires
Days 1-7 post-fires
MLS observations of this rare event
Enhanced CO between tropopause and 46 hPa seen for a month post-fires
Days 8-13 post-fires
Evidence of a filament of high CO heading west
These data could test large scale advection in UTLS, for CTMs or nudged CCMs.
Initial pyrocumulus convection associated with fire unlikely to be captured in reanalyses, but subsequent advection and mixing should be.
Days post-fires
Pumphrey et al., 2011
Days post-fires

17. Tracers: e.g., e90 to define tropopause
e90 tracer
Surface source, 90-day e-fold
Arbritrary emissions strength chosen to yield global mean steady state abundance of 100 ppb
Troposphere, on annual average = 80% of atmosphere
Using UCI CTM yields an e90 tropopause value of 90 ppb
White line e90 = 90 ppb
Prather et al 2011

18. Hand over to Sue…

19. TST from a Lagrangian perspective
METHOD: Ensembles of trajectories are initialised in the lower stratosphere and are integrated BACKWARD in time using ECMWF reanalysis winds and heating rates
Define the “tropospheric ensemble” by a TST criterion (e.g. trajectories must experience Pot. Temp. <340K within some time limit)
Trajectories can be traced through meteorological fields (e.g. TEMPERATURE) or other passive tracers and used to reconstruct high resolution observed fields
Lagrangian dry point
e.g. ADVECTION-CONDENSATION MODEL for stratospheric water vapour:
Water vapour at the endpoint of the trajectory is given by the Lagrangian Dry Point (LDP) (T, lat, lon, pres, time) of the tropospheric-ensemble: this is the last point at which trajectories encounter 100% RH before the endpoint.

20. Sensitivity of modelled H2O
Water vapour from TST-ensemble only
E4 - ERA-40
(3D-var assimilation)
EI - ERA-Interim
(4D-var assimilation)
Anomalously high values
in the extratropical lower stratosphere for ERA-40 trajectories
kin - kinematic
(vertical velocity
calculated from
continuity equation)
dia - diabatic
(vertical motion
from heating rates)
Liu et al., 2010

21. Diagnostic for TST - location of LDP
LDP occur mostly at the tropical tropopause (lowest T)
Large number of trajectories attain LDP near the extra-tropical tropopause in ERA-40 kinematic calculation
Liu et al., 2010

22. Defining the tropopause using trajectories
Transport statistics derived from trajectories can be used to give alternative definitions of the tropopause (e.g. Berthet et al. 2007)
Coloured contours show the proportion of backward trajectories that encountered the boundary layer (defined as log pressure height <1km) within the last 30 days. Solid line (20%) may be used to define the tropopause for all latitudes.
Significant differences from PV (white contours) especially in high latitudes and NH summer midlatitudes (influences of the Indian monsoon)
Liu, 2010 (PhD thesis)

23. Transport timescales - alternative view
For the inverse problem, here we initialise trajectories on 100hPa on 1x1 grids and integrate backwards for 5 years and study the distribution of TIME since last encountering the boundary layer (z* < 1km)
Note significant seasonal difference, especially role of the Tibetan plateau/Indian ocean in NH summer
We can convolute this with an idealised surface-emitting tracer to produce an estimate of its distribution on 100hPa based on transport alone, in a Lagrangian analogy of the e90 tracer (Prather et al. 2011)
Lifetime of the tracer will highlight transport features of corresponding timescales.

24. Summary
Identify a range of transport metrics, based on observational datasets
Test behaviour in UKCA and other UK models
Encourage use in wider model intercomparisons/evaluations (e.g. CCMI, HTAP)
This part of ACITES runs April 2013 for 2 years
We encourage your suggestions