First report on model evaluation Activity 3, WP 3.1.1

1
First report on model evaluationActivity 3, WP
3.1.1

• Christina Schnadt Poberaj and Johannes Staehelin
  
  
• ETH Zürich, Switzerland

2

• Participating models

• Three chemistry-transport models
• Oslo CTM2 (Gauss, UiO)
• p-TOMCAT (Dessens, UCAM-DCHEM)
• TM4 (Meijer, Van Velthoven, KNMI)
• Two chemistry-climate models (nudged mode)
• ECHAM5/MESSy (Jöckel, Hoor, MPI-CHEM)
• LMDz-INCA (Caro, Hauglustaine, LSCE)
• One chemistry-climate model (climate
  simulation)
  
• E39/C (Grewe, DLR)

3

• Evaluation of
• multi-model performance
• Simulations of selected years, comparison of
  model results with observations in the ETHmeg
  database
  
• Model evaluation year 2003 as part of the
  current impact study
  
• Campaign data to be used
• MOZAIC cruise and profile data, ozonesonde
  data,
  
• WDCGG surface observations, SPURT,CONTRACE II

4
ETHmeg Activity 3 website

• ETHmeg Activity 3 website contains
• useful information for modellers such as
  ozonesonde station coordinates, list of MOZAIC
  airports etc.
  
• instructions for model evaluation
• results graphically displayed

Model output
Validation against observations (restricted)
5

• Diagnostics of model evaluation

• Model evaluation set up of two parts
• Basic direct comparison of model results with
  observations including point-to-point
  comparison as in the TRADEOFF project (Brunner
  et al., 2003 2005)
• Model-to-model intercomparison of diagnostics
  describing key processes relevant for QUANTIFY
  

6
QUANTIFY model evaluation Ozonesonde stations
7
Comparison of modelled against ozonesonde
temperature
Lindenberg, Germany
8
Comparison of modelled against sonde ozone
Edmonton, Canada
LMDzINCA
Oslo CTM2
ECHAM5/MESSy
TM4
9
Summary Temperature

• Oslo CTM2 and TM4. Minor deviations from sonde
  temp- erature at most stations (? 2 K)
  
• ECHAM5/MESSy. Temperature generally somewhat
  lower than by sondes.
  
• Lower to middle troposphere, and lower
  stratosphere (LS) model temperatures 1-2 K
  less than by sondes (most stations)
  
• Upper troposphere (UT) model temperatures 3-5 K
  less than observed, a noticeable feature at all
  stations.
  
• LMDzINCA. Negative deviations in the troposphere
  of -1 K to -5 K maximising in the UT, and
  warm bias in the pressure range of 200 to 100
  hPa (up to 5 K).

10
Summary Ozone (1)

• Oslo CTM2.
• ? LS significant overestimation ( 50 )
• ? Troposphere, high latitudes, and midlatitude
  Canadian stations differences positive
  during winter and spring. During rest of the
  year, slight underestimation
• ? Troposphere, Europe more positive
  deviations
  
• ? Troposphere, Japan large overestimation of
  trop.strat. ozone.
  
• TM4.
• ? NH high and midlatitudes trop.strat.
  differences mostly relatively
  
• small and negative
• ? UT/LS maximum negative differences in the
  UT and maximum positive
  
• differences above in LS
• ? Lower stratosphere in subtropicstropics,
  Antarctica large overestimation
  
• ? Model ozone maybe higher than observed over
  Japan.

11
Summary Ozone (2)

• ECHAM5/MESSy.
• ? Mid- and high latitudes, NH and SH model
  ozone lower (higher) than observed in the
  troposphere (stratosphere)
  
• ? Negative trop. differences most pronounced in
  UT, effect largest at high latitudes
  
• ? Japan large positive differences in
  troposphere and stratosphere.
  
• LMDzINCA.
• ? LS, mid- and high latitudes, NH and SH model
  ozone -10 to -40 less than observed
  
• ? Troposphere Similar negative deviations at
  most times Exception summer, lower trop.,
  tendency to overestimate ozone
  
• ? UT, mid- and high latitudes, NH and SH model
  ozone larger than observations
  
• ? Japan, tropical stations, UT/LS large
  positive differences.
  
• All models large positive deviations of
  trop.strat. ozone from sondes
  
• over Japan
• ? Measure Japanese ozone sensors less ozone than
  other sensors?
  

12
Outlook what to do next?
Identify model biases Scatterplots O3, NOx, CO,

13
Testing the skill of the models Taylor diagrams

• correlation coefficient,
• pattern root-mean square (RMS) error,
• ratio of modelled/observed standard
• deviation
• all indicated by a single point

• Example C1
• correlation 0.52
• normalized standard deviation 1.2
• RMS error proportional to linear distance
  between Ref. and C1
  
• skill score 0.60

14
Identify troposphere-to-stratosphere transport
O3-CO correlations
From Hoor et al. (JGR, 2002)