upgrade&verification¬e:& evaluation&ofthe&e4suite&(fzpr

29
Grant agreement n°283576 MACCII Validation Subproject Report Upgrade verification note: Evaluation of the esuite (fzpr) for the period July 2013 January 2014 Date: January 2014 Lead Beneficiary: KNMI (#21) Nature: R Dissemination level: PU

Upload: others

Post on 18-Dec-2021

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

     

 

Grant  agreement  n°283576    

MACC-­‐II  Validation  Subproject  Report  

Upgrade  verification  note:  Evaluation  of  the  e-­‐suite  (fzpr)  for  the  period  July  2013  -­‐  January  2014    

Date:  January  2014  Lead  Beneficiary:  KNMI  (#21)  Nature:  R  Dissemination  level:  PU  

Page 2: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   2  /  29    

     

   

Work-­‐package   82  (VAL,  Validation)  Deliverable   -­‐  Title   Upgrade  verification  note:  

Evaluation  of  the  e-­‐suite  (fzpr),  for  the  period  July  2013  -­‐  January  2014  

Nature   R  Dissemination   PU  Lead  Beneficiary   KNMI  (#21)  Date   31  January  2014  Status   Final  Authors   A.  Benedictow  and  M.  Schulz  (MetNo),  A.  Blechschmidt  

and  A.  Richter  (IUP-­‐UB),  V.  Huijnen  (KNMI),  J.  Kapsomenakis  and  C.  Zerefos  (AA),  S.  Chabrillat,  Y.  Christophe,  B.  Langerock,  E.  Botek  (BIRA-­‐IASB),  M.  Razinger  (ECMWF),  A.  Wagner  (DWD)  

Editors   H.J.  Eskes  (KNMI),  V.  Huijnen  (KNMI)  Contact   info@copernicus-­‐atmosphere.eu    

     

This   document   has   been   produced   in   the   context   of   the   MACC-­‐II   project   (Monitoring   Atmospheric  Composition   and   Climate   -­‐   Interim   Implementation).   The   research   leading   to   these   results   has   received  funding   from   the   European   Community's   Seventh   Framework   Programme   (FP7   THEME   [SPA.2011.1.5-­‐02])  under  grant  agreement  n°  283576.  All  information  in  this  document  is  provided  "as  is"  and  no  guarantee  or  warranty  is  given  that  the  information  is  fit  for  any  particular  purpose.  The  user  thereof  uses  the  information  at   its   sole   risk   and   liability.   For   the   avoidance   of   all   doubts,   the   European   Commission   has   no   liability   in  respect  of  this  document,  which  is  merely  representing  the  authors  view.  

Page 3: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   3  /  29    

     

Summary  The  MACC  II  (Modelling  Atmospheric  Composition  and  Climate,  www.gmes-­‐atmosphere.eu)  project  is  establishing  the  core  global  and  regional  atmospheric  environmental  service  delivered  as  a  component  of  the  European  Earth  observation  programme  Copernicus  (previously  known  as  GMES,  Global  Monitoring  for  Environment  and  Security).  The  global  MACC  near-­‐real  time  (NRT)  service  provides  daily  analyses  and  forecasts  of  trace  gas  and  aerosol  concentrations.    

This  document  contains  verification  results  for  the  upgrade  of  the  NRT  service  planned  for  February  2014.  The  new  model  configuration  (the  e-­‐suite)  is  operated  in  parallel  to  the  operational  NRT  service  (the  o-­‐suite)  for  several  months.  For  more  details  about  the  validation  approaches  and  references  we  refer  to  the  NRT  validation  reports  of  MACC-­‐II.    

Below  the  main  results  are  summarised  from  a  comparison  of  the  performance  of  the  new  e-­‐suite  run  (fzpr,  period  July  2013  -­‐  January  2014),  the  operational  run  (o-­‐suite)  and  independent  observations.  Section  1  provides  a  brief  overview  of  the  changes  between  the  e-­‐suite  and  o-­‐suite.  Section  2  presents  the  validation  results.    

The  o-­‐suite  update  discussed  in  this  documents  consists  of  a  couple  of  relatively  minor  changes,  as  detailed  in  section  1.  The  impact  of  these  changes  on  the  performance  is  not  expected  to  be  very  large,  and  this  is  reflected  by  the  verification  results  summarised  below  and  shown  in  more  detail  in  section  2.  Similar  results,  or  improvements  (for  surface  ozone)  are  found  for  the  e-­‐suite.  We  find  a  notable  difference  in  polar  stratospheric  NO2,  where  the  new  e-­‐suite  is  an  improvement.  As  a  result  we  can  give  a  positive  advice  to  go  ahead  with  the  planned  upgrade  of  the  o-­‐suite.  

Aerosols  

As  compared  to  the  o-­‐suite,  an  initially  lower  correlation  against  Aeronet  NRT  data  is  found  in  July  and  August  2013,  possibly  due  to  spin-­‐up  time  deviations  from  an  optimal  state  of  the  model,  or  because  of  differences  originating  from  the  previous  upgrade  of  the  o-­‐suite,  which  took  place  on  7  October  2013.  In  September  through  to  November  2013  the  two  simulations  are  very  similiar  as  seen  in  the  temporal  evaluation  of  correlation.  The  same  applies  also  for  the  3  day  forecast  time  step  (o-­‐suite  72-­‐96h  vs  e-­‐suite  72-­‐96h).  The  statistics  for  November  2013  are  very  similar  (maybe  a  tiny  little  bit  better  for  the  e-­‐suite)  as  seen  in  the  scatterplots  against  Aeronet  AOD  observations  for  the  two  simulations.  In  view  of  the  large  similarity  it  is  probably  not  possible  (and  worth)  to  try  to  attribute  small  differences  in  aerosol  amount  and  bias  against  Aeronet  to  specific  changes  in  the  model.  No  other  reasons  could  be  found,  which  would  prevent  from  recommending  the  e-­‐suite  to  become  o-­‐suite.  

Hourly  dust  optical  depth  (DOD)  data  from  MACCII  e-­‐suite  has  been  evaluated  against  33  AERONET  level-­‐1.5  stations,  and  for  10  regions  (Western  Mediterranean,  Central  Mediterranean,  Eastern  Mediterranean,  North  Western  Maghreb,  Sahara,  Sahel,  Middle  East,  Subtropical  North  Atlantic,  Tropical  North  Atlantic  and  Middle  North  Atlantic)  for  the  period  August  1st  to  December  31st,  2013.  The  e-­‐suite  DOD  evaluation  shows  no  significant  differences  with  o-­‐suite.  It  may  be  emphasized  that  the  correlation  in  e-­‐suite  improves  slightly  in  some  regions,  like  the  Sahara,  the  Sahel  and  Middle  East,  but,  on  the  contrary,  a  slight  overestimation  of  DOD  is  observed  in  the  Sahara  and  Middle  East.  The  differences  between  e-­‐suite  and  o-­‐suite  are  small  and  hardly  significant.  

Ozone  

The  comparison  with  ozone  sonde  measurements  in  the  free  troposphere  shows  that  the  e-­‐suite  has  somewhat  lower  MNM  biases  in  all  regions  except  for  the  Arctic.  Especially  in  Antarctica,  e-­‐suite  shows  better  results.    

Page 4: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   4  /  29    

     

For  surface  ozone  mixing  ratios  at  the  GAW,  GMD  and  several  EMEP  stations,  the  e-­‐suite  shows  convincing  improvements  compared  to  o-­‐suite.  MNM  biases  are  mostly  lower  and  correlation  coefficients  higher,  except  for  the  stations  in  the  Southern  Hemisphere.      

It  should  be  noted  that  the  main  differences  are  observed  before  7  October  2013,  the  date  of  the  previous  o-­‐suite  update.  The  e-­‐suite  versus  o-­‐suite  differences  are  small  after  7  october,  demonstrating  the  good  agreement  compared  to  the  current  o-­‐suite.  

CO  

For  surface  CO,  compared  to  GAW  station  data,  e-­‐suite  shows  lower  MNM  biases  and  higher  correlation  coefficients  in  Asia  and  no  major  differences  to  o-­‐suite  compared  to  the  surface  observations  in  Europe.  For  the  two  Southern  Hemispheric  stations,  the  performance  of  o-­‐suite  is  in  parts  better.  

Tropospheric  NO2  

Model  validation  with  respect  to  GOME-­‐2  data  shows  that  tropospheric  NO2  columns  are  quite  well  reproduced  by  the  o-­‐suite  and  e-­‐suite,  indicating  that  emission  patterns  and  NOx  photochemistry  are  reasonably  represented.  However,  the  o-­‐suite  and  e-­‐suite  tend  to  underestimate  NO2  columns  over  land,  particularly  over  East-­‐Asia.  The  latter  may  result  from  an  underestimation  of  anthropogenic  NOx  emissions  in  the  inventories.  There  are  only  minor  differences  between  the  o-­‐suite  and  e-­‐suite.  

Formaldehyde  

Model  results  and  observations  are  in  good  agreement  for  HCHO  with  respect  to  magnitude.  There  is  almost  no  difference  between  the  o-­‐suite  and  e-­‐suite.  Satellite  values  are  generally  lower  than  model  values  for  Indonesia  where  biomass  burning  and  biogenic  sources  contribute  to  HCHO  emissions.  However,  this  is  not  the  case  for  North-­‐Africa  where  biomass  burning  and  biogenic  sources  contribute  to  HCHO  emissions  as  well.  Model  values  are  higher  than  satellite  values  for  East-­‐Asia.  

Stratospheric  ozone  

We  find  no  important  change  for  ozone  in  the  lower  stratosphere,  i.e.  in  the  region  which  contributes  most  to  UV  absorption.  The  e-­‐suite  delivers  up  to  10%  less  ozone  in  the  upper  stratosphere,  which  is  a  good  thing.  

The  validation  with  ozone  balloon  soundings  also  shows  that  there  are  no  major  differences  between  e-­‐suite  and  o-­‐suite.    

Stratospheric  Nitrogen  dioxide  (NO2)  

Both  e-­‐suite  and  o-­‐suite  catch  the  shape  of  time  series  of  stratospheric  NO2  columns  well  for  most  of  the  globe.  Nonetheless,  the  e-­‐suite  and  o-­‐suite  significantly  underestimate  the  magnitude  of  the  values  compared  to  satellite  observations.  However,  the  e-­‐suite  is  closer  to  satellite  observations  than  o-­‐suite  at  all  latitude  bands.    

A  significant  difference  in  stratospheric  NOx  is  found  above  the  South  Pole,  where  the  e-­‐suite  delivers  larger  concentrations  than  the  o-­‐suite  by  up  to  50%  in  the  lower  stratosphere.  This  difference  is  actually  due  to  some  error  in  the  previous  upgrade  of  the  o-­‐suite,  i.e.  on  2013-­‐10-­‐07  when  fnyp  was  upgraded  from  IFS  cycle  37R3  to  cycle  38R2:  a  discontinuity  shows  up  at  this  date  in  fnyp,  which  shows  afterwards  unusually  low  values  of  NOx  above  the  South  Pole.  These  unusually  low  values  are  still  present  4  months  after  the  date  of  the  upgrade.  

 

   

Page 5: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   5  /  29    

     

1 Configuration  of  o-­‐suite  and  e-­‐suite  

Key  model  information  is  given  on  the  MACC  pre-­‐operational  data-­‐assimilation  and  forecast  run  o-­‐suite  and  the  new  e-­‐suite.  Table  2.1  provides  information  on  the  satellite  data  used  in  the  o-­‐suite.    

1.1.1 o-­‐suite  

The  o-­‐suite  is  running  under  experiment  (EXPVER)  ‘fnyp’  and  is  described  in  Stein  et  al.  (2011)  and  references  therein.  Here  a  summary  of  the  main  physics  specifications  of  the  current  o-­‐suite  is  given.    

• The  meteorological  model  is  based  on  IFS  version  CY38R2,  see  also  http://www.ecmwf.int/products/data/technical/model_id/;  IFS  model  resolution  is  T255L60.  

• MOZART  version  3.5  is  used  (Kinnison  et  al.,  2007;  Emmons  et  al.,  2011)  with  a  resolution  of  1.125°x1.125°.  

• Anthropogenic  and  biogenic  emissions  are  based  on  the  MACCity  (Granier  et  al.,  2011)  and  (climatological)  MEGAN  emission  inventories  with  scaled  CO  emissions.  

• NRT  fire  emissions  are  taken  from  GFASv1.0  (Kaiser  et  al.  2012),  both  for  gas-­‐phase  and  aerosol.    

• Note  that  on  7  October  2013  the  o-­‐suite  has  been  upgraded  with  resulting  in  significant  modifications  to  the  chemical  analysis  system.  

 

Table  2.1:  Satellite  retrievals  of  reactive  gases  and  aerosol  optical  depth  that  are  actively  assimilated  in  the  o-­‐suite.  

Instrument   Satellite   Provider   Version   Type   Status  

MLS     AURA   NASA   V2  V3.4  

O3  Profiles  O3  Profiles  

20090901  -­‐  20130106  20130107  -­‐    

OMI     AURA   NASA   V883   O3  Total  column   20090901  -­‐  

GOME-­‐2   Metop-­‐A   Eumetsat   GDP  4.7   O3  Total  column   20131007  -­‐  

SBUV-­‐2   NOAA   NOAA   V8   O3  6  layer  profiles  O3  21  layer  profiles  

20090901  -­‐20131006  20131007-­‐  

IASI   MetOp-­‐A   LATMOS/ULB     CO  Total  column   20090901  -­‐  

MOPITT   TERRA   NCAR   V4  V5  

CO  Total  column  CO  Total  column  

20120705  -­‐  20130201  20130129  -­‐    

OMI   AURA   KNMI   DOMINO  

V2.0  

NO2  Tropospheric  column  

20120705  -­‐    

OMI   AURA   NASA   v003   SO2  Tropospheric  column  

20120705  -­‐    

MODIS   AQUA  /  TERRA  

NASA   Col.  5   Aerosol  total  optical  depth  

20090901  -­‐  

 

 

Page 6: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   6  /  29    

     

The  aerosol  model  The  aerosol  model  includes  12  prognostic  variables,  which  are  3  bins  for  sea  salt  and  desert  dust,  hydrophobic  and  hydrophilic  organic  matter  and  black  carbon,  sulphate  aerosols  and  its  precursor  trace  gas  SO2  (Morcrette  et  al.,  2009).  Aerosol  fire  emissions  are  based  on  GFASv1  (Kaiser  et  al.  2012).  A  variational  bias  correction  for  the  MODIS  AOD  is  in  place  based  on  the  approach  used  also  elsewhere  in  the  IFS  (Dee  and  Uppala,  2009).    

 

1.1.2 e-­‐suite    

The  e-­‐suite  with  experiment  ID  "fzpr"  has  been  running  for  the  period  starting  from  1  August  2013  and  is  now  running  in  near-­‐real-­‐time.  The  description  of  the  system  is  given  at:  http://www.copernicus-­‐atmosphere.eu/oper_info/global_system_changes/cy40r1/  

Here  a  summary  of  the  main  physics  changes  compared  to  the  current  o-­‐suite  is  given.    

• The  meteorological  model  is  based  on  IFS  version  CY40R1,  which  includes  amongst  others  modification  of  the  convection  to  address  diurnal  cycle  of  precipitation  and  changes  to  snow  albedo.  

• Global  forecasts  of  PM10  and  PM2.5  are  generated  

• Improved  bias  correction  for  MODIS  AOD  observations  

• Starting  from  21  January  2014  the  fire  emissions  have  been  updated  to  GFASv1.2  

• The  background  errors  have  been  rescaled  by  5%,  which  leads  to  a  small  reduction  of  the  weight  given  to  the  observations  in  the  assimilation.  

A  detailed  log  file  can  be  found  from:    http://www.copernicus-­‐atmosphere.eu/about/project_structure/global/g_idas/g_idas_2/e-­‐suite/    

 

   

Page 7: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   7  /  29    

     

2 Verification  results  for  the  e-­‐suite  "fzpr"  and  o-­‐suite    

2.1 Aerosol  

The  new  e-­‐suite  experiment  "fzpr"  has  been  evaluted  alongside  the  o-­‐suite  until  the  end  of  November  2013.  The  aerosol  optical  depth  comparisons  against  AERONET  are  shown  in  figures  2.1.1  to  2.1.3.  

 

 Fig.  2.1.1.  Mean  maps  of  AOD@550nm  for  Sep-­‐Nov  2013,  for  the  e-­‐suite  "fzpr"  and  o-­‐suite.    

 Fig.  2.1.2.  AOD@550nm  correlation  coefficient  o-­‐suite  and  e-­‐suite  model  simulation  against  Aeronet  NRT  level  1.5  data  for  Dec.  2011-­‐Nov  2013  (thick  red  curve);  Last  forecast  day  is  shown  separately  (light  red  curve).  

 

Page 8: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   8  /  29    

     

   

 Fig.  2.1.3.  Scatter  plot  of  AOD@550nm  for  the  e-­‐suite  "fzpr"  (top)  and  o-­‐suite  (bottom)  against  AERONET  ground  observations.    

 

Apart  from  aerosol  OD550,  we  have  studied  the  dust  optical  depth  in  the  Mediterranean  and  Sahara  regions.  We  have  assumed  as  Dust  optical  depth  (DOD)  as  the  AERONET  AOD  coarse  mode  obtained  from  the  spectral  shape  of  AOD  by  means  of  a  Spectral  Deconvolution  Algorithm  (SDA,  O’Neill  et  al.,  2003).  

DOD  MACCII  e-­‐suite  and  o-­‐suite  outputs  at  06,  09,  12,  15  and  18UTC  have  been  evaluated  with  near  DOD  observations  from  AERONET  averaged  for  these  hours  (±1.5h)  at  the  33  AERONET  stations.  Skill  scores  have  been  computed  for  each  station  on  a  daily  basis  and  averaged  for  the  10  geographical  regions.  Results  are  given  below  

 

Page 9: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   9  /  29    

     

 Figure  2.1.4:  Map  of  AERONET  level-­‐1.5  stations  used  in  this  analysis.  

 

  DOD  MACCII  o-­‐suite   DOD  MACCII  e-­‐suite  Region   MB   FGE   RMSE   r   MB   FGE   RMSE   r   n  

Western  Mediterranean   -­‐0.03   0.84   0.10   0.50   -­‐0.02   0.89   0.11   0.50   2720  

Central  Mediterranean   -­‐0.03   0.95   0.18   0.37   -­‐0.03   0.98   0.18   0.38   420  

Eastern  Mediterranean   -­‐0.02   0.75   0.08   0.53   -­‐0.02   0.85   0.08   0.54   2360  

Middle  North  Atlantic   -­‐0.03   1.09   0.10   0.38   -­‐0.04   1.19   0.10   0.41   1494  

Subtropical  North  Atlantic   -­‐0.01   0.87   0.24   0.45   -­‐0.01   0.95   0.24   0.48   456  

Tropical  North  Atlantic   0.00   0.41   0.15   0.63   0.01   0.41   0.16   0.65   337  

North  Western  Maghreb   -­‐0.01   0.71   0.13   0.42   -­‐0.01   0.70   0.13   0.45   895  

Sahara   0.08   0.86   0.30   0.33   0.14   0.93   0.31   0.43   526  

Sahel   -­‐0.02   0.51   0.30   0.35   -­‐0.01   0.51   0.30   0.38   1455  

Middle  East   0.05   0.40   0.11   0.72   0.08   0.43   0.14   0.78   2234  

 Table  2.1.1.  The  main  skill  scores,  averaged  by  regions.  

 

   Figure  2.1.5.  Some  representative  graphics  of  the  e-­‐suite  evaluation  for  the  10  regions.  

Page 10: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   10  /  29    

   

   

   

   

     Figure  2.1.5  (continued).  Some  representative  graphics  of  the  e-­‐suite  evaluation  for  the  10  regions.  

Note:  A  detailed  analysis  has  been  performed  for  each  station.  This  information  is  available  if  required.  

 

Page 11: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   11  /  29    

   

2.2 Tropospheric  Ozone  

2.2.1 Validation  with  sonde  data  in  the  free  troposphere  

In  the  free  troposphere,  e-­‐suite  has  lower  MNM  biases  except  for  the  Arctic  in  September  and  October.  In  Antarctica  in  August  and  September,  the  MNMB  is  significantly  lower  for  e-­‐suite,  see  Figure  2.2.2.    

 Figure  2.2.1:  Modified  normalized  mean  bias  (%)  of  ozone  of  o-­‐suite(solid)  and  e-­‐suite  (dotted)  model  runs  against  aggregated  sonde  data  in  the  free  troposphere  in  3  regions.  

 Figure  2.2.2.  O3  partial  pressures,  sonde  (red/green)  compared  to  model  (black),  for  o-­‐suite  (left)  and  e-­‐suite  (right)  over  Neumayer  station  for  24.08.2013.    

Free Troposphere 2013

-40

-30

-20

-10

0

10

20

30

40

50

Aug 2013 Sep 2013 Oct 2013 Nov 2013

Month

MN

MB

[%]

MACC_osuite_Antarctica

MACC_osuite_NorthernMidlatitudes

MACC_osuite_Arctic

MACC_esuite_Antarctica

MACC_esuite_Arctic

MACC_esuite_NorthernMidlatitudes

Page 12: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   12  /  29    

   

 Figure  2.2.3:  Modified  normalized  mean  bias  in  %  (left)  and  correlation  coefficients  (right)  of  the  o-­‐suite  (red)  and  e-­‐suite  (orange)  runs  compared  to  observational  O3  surface  GAW  data  between  September  –November  2013    

 Table  2.2.1:  MNM  biases  in  %  and  correlation  coefficients  of  o-­‐suite  and  e-­‐suite  for  O3  

2.2.2 Validation  with  GAW  in-­‐situ  data  

For  European  stations  (HPB,  JFJ,  ZUG,  SON,  MCI),  e-­‐suite  (fzpr)  shows  MNM  biases  between  -­‐8  and  35%,  while  o-­‐suite  has  MNM  biases  between  4  and  38%;  at  4  of  5  stations  e-­‐suite  shows  an  improvement  in  MNM  bias  compared  to  o-­‐suite,  see  Fig  s  and  3.  Correlation  coefficients  for  e-­‐suite  are  mostly  slightly  higher,  see  Fig  2.2.3  and  Table  2.2.1.    

For  the  three  Asian  GAW  stations  (RYO,  YON,  MNM),  e-­‐suite  shows  a  slight  reduction  of  the  modelled  O3  overestimation  with  an  improvement  in  MNM  biases  (for  e-­‐suite  MNM  Biases  between  5  and  32%,  for  o-­‐suite  between  8  and  34%),  see  Fig.  2.2.6.  The  correlation  shows  an  improvement  for  e-­‐suite.    

For  southern  hemispheric  stations,  O3  mixing  ratios  are  generally  lower  for  e-­‐suite,  which  leads  to  a  reduction  of  the  MNM  bias  for  Cape  Point  station  and  partly  to  an  underestimation  for  the  other  two  stations  located  in  the  Southern  Hemisphere,  see  time  series  plot  in  Fig.  2.2.7  and  2.2.8.  MNM  biases  are  between  -­‐10  and  8  %  for  e-­‐suite  and  between  20  and  31%  for  o-­‐suite,  correlation  coefficients,  however,  are  lower  for  e-­‐suite.  

 Figure  2.2.4:  Time  series  for  o-­‐suite  (left)  and  e-­‐suite  (right)  compared  to  GAW  observations  (blue  dots)  at  Sonnblick  station  between  September  and  November  2013.      

O3 September to November 2013

-20

0

20

40

60

HPBJF

JZUG

SONMCI

RYOYON

MNMCVO

CPTUSH

NEU

Station (GAW)

MN

MB

[%]

MACC_osuite(fnyp)

MACC_esuite(fzpr)

Europe Asia

Southern hemispherehemisphere

Cape Verde

O3 September to November 2013

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

HPBJF

JZUG

SONMCI

RYOYON

MNMCVO

CPTUSH

NEU

Station (GAW)

Cor

rela

tion

Coe

ffici

ent r

MACC_osuite(fnyp)MACC_esuite(fzpr)

Europe AsiaSouthern hemisphere

Cape Verde

O3 Sep-Nov 2013 HPB JFJ ZUG SON MCI RYO YON MNM CVO CPT USH NEUMNMB MACC_esuite (fzpr) 34.7 -1.0 16.3 8.3 -8.5 5.4 7.8 31.5 18.0 8.3 -10.1 6.0MNMB MACC_osuite (fnyp) 38.1 4.1 21.8 13.2 8.5 8.3 9.2 34.2 20.1 19.5 31.0 25.5R MACC_esuite (fzpr) 0.63 0.59 0.32 0.53 0.55 0.35 0.54 0.75 0.75 0.59 0.50 -0.16R MACC_osuite (fnyp) 0.59 0.54 0.31 0.52 0.58 0.29 0.50 0.69 0.70 0.73 0.66 0.49

Page 13: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   13  /  29    

   

 

 Figure  2.2.5:  Time  series  for  o-­‐suite  (left)  and  e-­‐suite  (right)  compared  to  GAW  observations  (blue  dots)  at  Monte  Cimone  station  between  September  and  November  2013.      

 

 Figure  2.2.6:  Time  series  for  o-­‐suite  (left)  and  e-­‐suite  (right)  compared  to  GAW  observations  (blue  dots)  at  Minamitorishima  station  between  September  and  November  2013.      

 

 Figure  2.2.7:  Time  series  for  o-­‐suite  (left)  and  e-­‐suite  (right)  compared  to  GAW  observations  (blue  dots)  at  Cape  Point  station  between  September  and  November  2013.      

Page 14: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   14  /  29    

   

 Figure  2.2.8:  Time  series  for  o-­‐suite  (left)  and  e-­‐suite  (right)  compared  to  GAW  observations  (blue  dots)  at  Neumayer  station  between  September  and  November  2013.        

2.2.3 Validation  with  in-­‐situ  station  data  (AA)  

For  the  e-­‐suite  validation,  simulated  surface  O3  concentrations  are  compared  against  observed  surface  O3  at  a  number  of  stations  listed  in  Table  2.2.2.  The  validation  results  span  over  the  five-­‐month  period  August-­‐December  2013.  Time-­‐series  and  validation  scores  are  shown  in  the  following  figures  2.2.9  -­‐  2.2.14.  All  the  left-­‐column  figures  show  the  validation  with  the  new  model  version  (e-­‐suite,  fzpr)  and  on  the  right  the  validation  with  the  previous  model  version  (o-­‐suite,  fnyp).  Figure  2.2.9  shows  validation  results  at  Finokalia  (Crete)  in  the  eastern  Mediterranean.  Figure  2.2.10  provides  similar  information  for  the  US,  Figure  2.2.11  for  the  tropics,  Figure  2.2.12  for  Lauder  (New  Zealand)  and  Figures  2.2.13  and  2.2.14  for  the  Arctic  and  Antarctic.    

The  validation  results  can  be  summarized  as  follows:  At  Finokalia  station  both  models  reproduce  well  mean  surface  ozone  concentrations  (bias≈0%)  as  well  as  the  day  by  day  variability  (correlation  between  simulated  and  observed  daily  surface  ozone  concentrations  r=0.8).  Over  USA  stations  both  e-­‐suite  and  o-­‐suite  runs  overestimate  ozone  mixing  ratios.  However,  e-­‐suite  run  shows  lower  biases  than  o-­‐suite  in  all  USA  stations.  Over  tropical  stations  (BAR,  BER,  MLO  and  SMO)  both  e-­‐suite  and  o-­‐suite  runs  overestimate  ozone  mixing  ratios.  Again  e-­‐suite  shows  a  better  performance  in  bias  than  o-­‐suite.  Correlations  between  simulated  and  observed  surface  ozone  concentrations  are  high  for  both  runs.  At  Lauder  station  (New  Zealand)  e-­‐suite  reproduces  well  mean  surface  ozone  concentrations  (bias≈0%)  while  o-­‐suite  overestimates  it  by  7  ppb.  The  e-­‐suite  is  slightly  better  at  reproducing  the  day  by  day  variability  at  Lauder.  Finally  e-­‐suite  run  shows  lower  biases  (in  absolute  values)  also  over  Arctic  and  Antarctica  stations.  To  summarize  e-­‐suite  run  performers  better  than  o-­‐suite  over  all  continents  in  terms  of  bias.  

   

Page 15: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   15  /  29    

   

 

Mediterranean  

 

Figure  2.2.9:  Time  series  for  e-­‐suite  (fzpr,  left)  and  o-­‐suite  (fnyp,  rigth)  compared  to  Mediterranean  stations  observed  ozone  values  (blue  dots)  between  August  and  December  2013  

USA  

 

 

Figure  2.2.10:  Time  series  for  e-­‐suite  (fzpr,  left)  and  o-­‐suite  (fnyp,  rigth)  compared  to  USA  stations  observed  ozone  values  (blue  dots)  between  August  and  December  2013  

Page 16: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   16  /  29    

   

Tropics  

 

 

Figure  2.2.11:  Time  series  for  e-­‐suite  (fzpr,  left)  and  o-­‐suite  (fnyp,  rigth)  compared  to  Tropical  stations  observed  ozone  values  (blue  dots)  between  August  and  December  2013  

Page 17: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   17  /  29    

   

Figure  2.2.12:  Time  series  for  e-­‐suite  (fzpr,  right)  and  o-­‐suite  (fnyp,  left)  compared  to  Lauder  (45.04°S,  169.68°E)  observed  ozone  values  (blue  dots)    between  August  and  December  2013.  

Arctic  

Figure  2.2.13:  Time  series  for  e-­‐suite  (fzpr,  left)  and  o-­‐suite  (fnyp,  rigth)  compared  to  Arctic  stations  observed  ozone  values  (blue  dots)  between  August  and  December  2013  

   

Page 18: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   18  /  29    

   

Antarctica  

 

Figure  2.2.14:  Time  series  for  e-­‐suite  (fzpr0,  right)  and  o-­‐suite  (fnyp,  left)  compared  to  Antarctica  stations  observed  ozone  values  (blue  dots)  between  August  2013  and  December  2013  

Table  2.2.2:  Coordinates  of  stations  used  in  the  present  validation  analysis.  

Station Latitude Longitude Altitude (m) Country Summit (SUM) 72.57°N 38.38°W 3266 Greenland Barrow (BRW) 71.32°N 156.61°W 8 Alaska, United States Moody (WKT) 31.32°N 97.33°W 260 Texas, United States Boulder Atmospheric Observatory (BAO)

40.05°N 105.00°W 1584 Colorado, United States

Trinidad Head (THD) 41.05°N 124.15°W 107 California, United States Finokalia (FK) 35.32°N 25.67°E 250 Greece Ragged Point (BAR) 13.17°N 59.46°W 45 Barbados Bermuda (BER) 32.27°N 64.88°W 30 United Kingdom Mauna Loa (MLO) 19.54°N 155.58°W 3397 Hawaii, United States Tutuila (SMOC) 14.23°S 170.56°W 77 American Samoa Lauder (LDR) 45.04°S 169.68°E   370 New Zealand Arrival Heights (ARH) 77.80°S 166.78°W 50 New Zealand, Antarctica South Pole (SPO) 90.00°S 24.80°W 2837 Antarctica

Page 19: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   19  /  29    

   

2.3 Carbon  monoxide  (CO)  surface  concentrations  

For  European  stations,  there  is  only  marginal  difference  in  MNM  bias  between  MACC  _e-­‐suite  and  o-­‐suite.  Both  model  runs  show  the  same  underestimation  of  CO  mixing  ratios.  Please  note  that  for  Zugspitze  station,  CO  observational  data  is  biased  due  to  calibration  problems  during  the  evaluation  time  period.  e-­‐suite  shows  slightly  higher  correlation  coefficients  than  o-­‐suite.    

For  the  three  Asian  GAW  stations,  e-­‐suite  corresponds  better  with  the  observations  and  shows  lower  MNM  biases  and  mostly  higher  correlation  coefficients,  see  Fig.  2.3.1  and  2.3.4.    

For  the  Southern  Hemispheric  station  Ushuaia,  the  performance  of  the  both  model  runs  is  mostly  identical,  however,  the  correlation  coefficient  is  negative  for  e-­‐suite.  For  Cape  point  station  e-­‐suite  shows  a  slightly  higher  MNM  bias  and  an  identical  correlation.  For  Cape  Verde  station,  e-­‐suite  shows  improved  results,  especially  for  September,  see  Fig.  2.3.3.    

 Figure  2.3.1:  Modified  normalized  mean  bias  in  %  of  the  o-­‐suite  (red)  and  e-­‐suite  (orange)  runs  compared  to  observational  CO  surface  GAW  data.  

 Figure  2.3.2:  Time  series  for  o-­‐suite  (left)  and  e-­‐suite  (right)  compared  to  GAW  observations  (blue  dots)  at  Jungfraujoch  station  between  September  and  November2013.      

 Figure  2.3.3:  Time  series  for  o-­‐suite  (left)  and  e-­‐suite  (right)  compared  to  GAW  observations  (blue  dots)  at  Cape  Verde  station  between  Septemberand  November  2013.    

CO September to November 2013

-50.0-40.0-30.0-20.0-10.0

0.010.020.030.040.050.0

HPBJF

JZUG

SONMCI

RYOYON

MNMCVO

CPTUSH

Station (GAW)

MN

MB

[%]

MACC_osuite(fnyp)

MACC_esuite(fzpr)

Europe Asia

Southern hemisphere

Cape Verde

observational data too high

CO September to November 2013

0.000.100.200.300.400.500.600.700.800.901.00

HPBJF

JZUG

SONMCI

RYOYON

MNMCVO

CPTUSH

Station (GAW)

Cor

rela

tion

Coe

ffici

ent r

MACC_osuite(fnyp)

MACC_esuite(fzpr)

Europe Asia

Southern hemisphere

Cape Verde

Page 20: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   20  /  29    

   

   Figure  2.3.4:  Time  series  for  o-­‐suite  (left)  and  e-­‐suite  (right)  compared  to  GAW  observations  (blue  dots)  at  Minamitorishima  station  between  September  and  November  2013.      

 

 Table  2.3.1:  MNM  biases  in  %  and  correlation  coefficients  of  o-­‐suite  and  e-­‐suite  for  September  to  November  2013  for  CO.  

The  scorecard  shown  in  figure  2.3.5  confirms  the  relatively  small  changes  between  o-­‐suite  and  e-­‐suite  for  ozone  and  CO,  with  also  little  dependence  on  the  forecast  step.  

 

 

CO Sep-Nov 2013 HPB JFJ ZUG SON MCI RYO YON MNM CVO CPT USHMACC_esuite (fzpr) -19.6 -16.7 -27.4 -20.6 -12.7 -3.2 -5.2 -1.9 -3.9 26.0 -10.1MACC_osuite (fnyp) -18.8 -17.4 -27.8 -23.0 -13.1 -5.0 -7.2 -4.0 -7.7 19.2 -11.2MACC_esuite (fzpr) 0.61 0.72 0.51 0.62 0.21 0.70 0.71 0.75 0.65 0.42 -0.08MACC_osuite (fnyp) 0.60 0.66 0.46 0.61 0.20 0.66 0.73 0.76 0.56 0.44 0.17

Page 21: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   21  /  29    

   

 Figure  2.3.5:  Scorecard  for  the  comparison  of  the  e-­‐suite  and  o-­‐suite  performance  with  respect  to  GAW  station  observations  of  ozone  (top  part)  and  CO  (bottom  part).  Shown  are  the  normalised  bias  (MNMB),  fractional  gross  error  (FGE)  and  correlation  for  the  different  GAW  stations.  Green  colors  indicate  a  better  performance  of  the  new  e-­‐suite  (fwu0)  and  pink  colors  a  worse  performance.  The  hatched  (densely  hatched)  areas  show  datasets  with  a  poor  data  availability  of  less  than  50%  (25%).  The  horizontal  axis  is  the  forecast  step  (0-­‐96h).  

 

 

   

Page 22: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   22  /  29    

   

2.4 Tropospheric  Nitrogen  dioxide  (NO2)  

Model  validation  with  respect  to  GOME-­‐2  data  shows  that  tropospheric  NO2  columns  (Figure  2.4.1)  are  quite  well  reproduced  by  the  o-­‐suite  and  e-­‐suite,  indicating  that  emission  patterns  and  NOx  photochemistry  are  reasonably  represented.  However,  the  o-­‐suite  and  e-­‐suite  tend  to  underestimate  NO2  columns  over  land,  particularly  over  East-­‐Asia.  The  latter  may  result  from  an  underestimation  of  anthropogenic  NOx  emissions  in  the  inventories.  There  are  only  minor  differences  bewteen  the  o-­‐suite  and  e-­‐suite.  

 

   

Figure  2.4.1:  Time  series  of  average  tropospheric  NO2  columns  [1015  molec  cm-­‐2]  from  GOME-­‐2  compared  to  model  results  for  different  regions.  

   

Page 23: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   23  /  29    

   

2.5 Formaldehyde  (HCHO)  

Model  results  and  observations  are  in  good  agreement  for  HCHO  (see  Figure  2.5.1)  with  respect  to  magnitude.  There  is  almost  no  difference  between  the  o-­‐suite  and  e-­‐suite.  Satellite  values  are  generally  lower  than  model  values  for  Indonesia  where  biomass  burning  and  biogenic  sources  contribute  to  HCHO  emissions.  However,  this  is  not  the  case  for  North-­‐Africa  where  biomass  burning  and  biogenic  sources  contribute  to  HCHO  emissions  as  well.  Model  values  are  higher  than  satellite  values  for  East-­‐Asia.  

 

   Figure  2.5.1:  Time  series  of  average  tropospheric  HCHO  columns  [1016  molec  cm-­‐2]  from  GOME-­‐2  compared  to  model  results  for  different  regions.  

 

   

Page 24: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   24  /  29    

   

2.6 Stratospheric  ozone  evaluation  

In  the  stratosphere  the  validation  with  balloon  sondes  revealed  that  there  are  no  major  differences  between  e-­‐suite  (fzpr)  and  o-­‐suite  (fnyp),  see  Figure  2.6.1    

 Figure  2.6.1:  Modified  normalized  mean  bias  (%)  of  ozone  of  o-­‐suite  /  e-­‐suite  (dotted  line)  model  runs  against  aggregated  sonde  data  in  the  stratosphere.  

Quick  looks  (e.g.  by  the  comparison  tool  at  the  stratospheric  ozone  service  –  see  Fig.  2.6.2)  show  no  difference  between  the  e-­‐suite  and  o-­‐suite.  Fig.  2.6.3  shows  the  relative  difference  as  a  function  of    latitude  and  pressure  for  2013-­‐12-­‐18  12:00  (similar  results  for  other  dates).  We  see  that  the  difference  never  exceeds  5%  between  10  hPa  and  the  tropopause  (most  important  region).  In  the  upper  stratosphere  the  e-­‐suite  analysis  delivers  up  to  10%  less  ozone  than  the  o-­‐suite  analysis.  This  is  a  good  thing  since  the  validation  of  o-­‐suite  against  ozonesondes,  for  period  2009-­‐09  to  2012-­‐09  (experiment  f93i)  shows  a  slight  overestimation  at  all  pressures  smaller  than  20  hPa.  

 

 Figure  2.6.2:  Examples  of  stratospheric  ozone  map  by  the  e-­‐suite  (left)  and  o-­‐suite  (right):    50hPa,  26  January  2014.  

Stratosphere 2013

-40

-30

-20

-10

0

10

20

30

40

50

Aug 2013 Sep 2013 Oct 2013 Nov 2013

Month

MN

MB

[%]

MACC_osuite_Antarctica

MACC_osuite_NorthernMidlatitudes

MACC_osuite_Arctic

MACC_esuite_Antarctica

MACC_esuite_Arctic

MACC_esuite_NorthernMidlatitudes

Page 25: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   25  /  29    

   

 

   Figure  2.6.3:  Relative  difference  of  ozone  by  the  o-­‐suite  and  e-­‐suite  (e-­‐o/o)  on  2013-­‐12-­‐18.  Left  pane:  zonally  averaged  differences;  right  pane:  extrema  of  the  differences  encountered  on  each  zonal  circle.  Green  dots  show  the  location  of  the  (PV-­‐based)  tropopause.  

The  similarity  between  o-­‐suite  and  e-­‐suite  is  confirmed  by  comparisons  with  ground-­‐based  observations  as  provided  by  the  EU  project  NORS.  Fig.  2.6.4  shows  the  stratospheric  column  (25-­‐60km)  above  Ny-­‐Alesund  as  a  function  of  time.  The  partial  column  gives  much  more  weight  to  the  lower  stratosphere,  and  the  e-­‐suite  results  are  nearly  undistinguishable  from  the  o-­‐suite  results.  The  next  VAL  3-­‐monthly  report  will  discuss  the  systematic  bias  found  in  both  cases  with  the  ground-­‐based  MWR  instrument.  

 

   

Figure  2.6.4:  Time  series  of  Ozone  stratospheric  columns  observed  by  the  MWR  instrument  above  Ny  Alesund,  and  corresponding  results  by  o-­‐suite  (full  red  line)  and  e-­‐suite  (dashed  red  line).  Comparison  provided  by  EU  FP7  project  NORS.  

   

Page 26: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   26  /  29    

   

2.7 Stratospheric  Nitrogen  dioxide  (NO2)  

Both  runs  catch  the  shape  of  time  series  of  stratospheric  NO2  columns  (see  Figure  3)  well.  Nonetheless,  the  e-­‐suite  and  o-­‐suite  significantly  underestimate  the  magnitude  of  the  values  compared  to  satellite  observations.  However,  e-­‐suite  results  are  closer  to  satellite  observations  than  o-­‐suite  results  at  all  latitude  bands.    

 

 Figure  2.7.1:  Time  series  of  average  stratospheric  NO2  columns  [1015  molec  cm-­‐2]  from  GOME-­‐2  compared  to  model  results  for  different  latitude  bands.  

   

Page 27: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   27  /  29    

   

 

Figure  2.7.2:  Examples  of  stratospheric  map  of  the  NOx  mass  mixing  ratio  by  the  e-­‐suite  (left)  and  o-­‐suite  (right):  50hPa,  26  January  2014.  

Figure  2.7.3:  NOx  mass  mixing  ratio  by  the  o-­‐suite  at  50hPa  on  6,  7  and  8  October  2013  

Quick  looks  show  no  important  difference  between  the  NOx  delivered  by  the  e-­‐suite  and  the  o-­‐suite  in  the  lower  stratosphere,  except  above  the  South  Pole  where  an  unexpectedly  large  difference  is  found:  fig.  2.7.2  shows  that  at  50  hPa  for  the  end  of  January,  the  e-­‐suite  delivers  max  3.1  ppbm,  i.e.  35%  more  NOx  than  the  2.3  ppbm  in  the  o-­‐suite  analysis.  

Exploring  for  other  dates  and  projections,  we  find  that  this  difference  appeared  on  2013-­‐10-­‐07,  i.e.  when  the  o-­‐suite  was  upgraded  from  the  previous  e-­‐suite  (fwu0).  Fig.  2.7.3  shows  that  a  discontinuity  appears  on  that  date  in  the  o-­‐suite  while  there  is  no  such  discontinuity  in  the  e-­‐suite.    

The  time-­‐series  of  NOx  mmr  at  50  hPa  by  the  consolidated  o-­‐suite  (i.e.  f93i+fnyp)  confirms  that  for  Antarctic  spring  post-­‐20131007,  the  low  values  found  by  fnyp  above  the  South  Pole  are  quite  unusual  (fig.2.7.4).  

 

Page 28: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   28  /  29    

   

 

Figure  2.7.4:  NOx  mass  mixing  ratio  by  the  e-­‐suite  at  50hPa  on  6,  7  and  8  October  2013  

 

Figure  2.7.5:  Time-­‐latitude  plot  of  NOx  mass  mixing  ratio  by  the  o-­‐suite  at  50hPa.  The  oval  in  brown  shows  the  January  2011  period  discussed  below;  the  oval  in  cyan  shows  the  unusually  low  values  above  the  South  Pole  during  Antarctic  Spring  2013.  

It  is  unfortunately  not  possible  to  compare  directly  the  model  output  with  (vertically  resolved)  observations  of  NO2,  because  the  observations  of  NO  and  NO2  by  ACE-­‐FTS  are  not  available  after  2012-­‐08.  So  we  have  evaluated  the  o-­‐suite  output  for  January  2011  (fig.  2.7.6).  This  shows  that  the  larger  values  for  that  year  are  correct  (in  the  lower  stratosphere),  so  the  lower  values  for  January  2014  are  most  probably  underestimated.  

 

Page 29: Upgrade&verification¬e:& Evaluation&ofthe&e4suite&(fzpr

File: MACCII_VAL_e-suite_20140131.docx

   29  /  29    

   

 

Figure  2.7.6:  Vertical  profile  of  NOx  mass  mixing  ratio  averaged  in  latitude  band  90°S-­‐60°S  for  2011-­‐01,  as  observed    by  ACE-­‐FTS  (dots)  and  -­‐  after  interpolation  to  these  obs  -­‐  delivered  by  the  o-­‐suite  (i.e.  f93i  ;  red  line)  and  BASCOE  offline  experiment  (eb0135A;  note  NO  and  NO2  are  not  constrained).