Observa(on-­‐Driven  Studies  Using  GEOS-­‐5  Earth  System  Modeling  and  

Analysis:  Some  Examples  

Steven  Pawson  Global  Modeling  and  Assimila(on  Office  Earth  Sciences  Division,  NASA  GSFC  

Transla(ng  Process  Understanding  to  Improve  Climate  Models:  Princeton,  NJ,  10/15-­‐16/2015  

GMAO:  Themes  

Weather  Analysis  and  Predic(on  

Seasonal-­‐to-­‐Decadal  Analysis  and  Predic(on  

Global  Mesoscale  Modeling   Reanalysis  Observing  System  

Science  

•  These  (non-­‐orthogonal)  themes  decribe  GMAO’s  main  ac(vi(es  •  GEOS-­‐5  is  a  modular  system,  encompassing  much  of  the  Earth  System  •  All  themes  include  mul(ple  components  of  the  Earth  System    •  Strong  emphasis  on  NASA’s  Earth  Observa(ons        

GEOS-­‐5:  An  Overview  Modular  Earth  System  Model:    •  Atmosphere,  land,  ocean,  cryosphere  (physics,  chemistry,  biology)  •  Aerosols,  chemistry,  carbon  cycle,  …    •  Early  adopter  of  a  modular  infrastructure  (ESMF)  •  Used  at  resolu(ons  of  about  100km  to  a  few  km  Data  Assimila(on:    •  Atmospheric  assimila(on  developed  jointly  with  NCEP    •  Ocean  and  land  are  developed  in  house  Observa(ons:    •  Use  a  broad  range  of  NASA  and  non-­‐NASA  data    •  Ability  to  synthesize  observing  systems  from  model  fields  

Sea-­‐Ice  Predic(on  from  Seasonal  Forecasts  GEOS-­‐5  Mid-­‐May  Ensemble:  Valid  September  2014   GSFC  SSMIS  –  DMSP-­‐F17  September  2014  

Relevance  for  this  Mee(ng    

Data  Assimila(on:    •  confronta(on  of  the  GEOS-­‐5  model  with  observa(ons  is  the  central  

theme  of  GMAO’s  work      Spa(o-­‐temporal  structure:    •  the  broad  range  of  (me  (hours  to  decades)  and  space  (few-­‐km  to  

100-­‐km  grids)  demands  drives  GMAO  to  use  resolu(on-­‐aware  parameteriza(ons    

Example  1:  Surface  Drag  over  the  Ocean  

Surface  wind  speed  versus  zo  over  ocean    A  change  to  the  func(onal  rela(onship  between  ocean  roughness  and  wind  stress  was  introduced  into  GEOS-­‐5.      This  was  based  on  oceanic  observa(ons  (Edson,  2011)  implying  that  the  drag  in  GEOS-­‐5  was  too  weak.          Objec(ve  was  to  improve  surface  wind  speeds  over  the  oceans    Work  led  by  Andrea  Molod  

Impact  on  Simulated  Surface  Winds  Surface Wind Speed m/sec

MODEL  

GSSTF  

Observa(on-­‐derived  surface  wind  speed  

Old  Model  minus  Observa(ons   New  Model  minus  Observa(ons  

The  increased  surface  roughness  over  oceans  leads  to  a  substan(al  improvement  in  surface  winds  in  the  DJF  season  (impact  extends  into  the  stratosphere).    [Garfinkel  et  al.,  2013)    

An  Addi(onal  Improvement  

Surface  wind  speed  versus  zo  over  ocean    

Z0  (m

)  

Wind  Speed  (m/sec)  

Few  observa(ons  of  surface  roughness  exist  in  the  vicinity  of  tropical  storms.    In  GEOS-­‐5,  the  roughness  increased  with  wind  speed,  reaching  unrealis(c  values.    An  upper  bound  was  imposed,  following  work  by  Emanuel  (1996,  2003),  Donelan  (2004),  and  others.        

Impact:  Typhoon  Megi  in  2010  was  more  realis(c  with  capped  drag  (Molod,  2012)        

GEOS-­‐5  AGCM:  Resolu(on-­‐Aware  Cumulus  Convec(on    Approach  adopted  by  Molod  et  al.  (2015)  in  GEOS-­‐5  builds  on  the  “Stochas(c  Tokioka”  model  by  using  introducing  a  limit  that  depends  on  resolu(on  of  the  GCM.  (See  also  Lim  et  al.,  2015.)        

PDF of Minimum Allowable Entrainment  “Stochas(c  Tokioka”  (Bacmeister  and  Stephens,  2011)  describes  the  use  of  a  variable  Tokioka  (1988)  parameter,  which  places  a  minimum  on  the  entrainment  into  deep  convec(ve  clouds.  Tallest  (non-­‐entraining)  large  mass  flux  convec(ve  events  are  eliminated,  and  total  cumulus  mass  flux  for  deep  convec(on  is  restricted.  Values  are  sampled  from  a  PDF  with  prescribed  parameters).  

Impact  on  resolved/parametrized  precipita(on    

~25  km  ~50  km  

~100  km  Total  Conv+Anvil  LargeScale  

~200  km  

Improved  Tropical  Cyclones  in  MERRA-­‐2  

Even through the spatial resolution of MERRA-2 is not much better than that of MERRA (close to 50km), this plot shows that the hurricane is developed much more realistically in MERRA-2. Largely a consequence of the model improvements. (This is a height section through the core.)

Category-5 Hurricane David (1979) caused widespread damage from the Caribbean to the U.S.. It was virtually undetected in models of the day. It is one of several historical hurricanes in MERRA-2 that is realistically depicted for the first time in a gridded data set.

MERRA   MERRA-­‐2  

Diagnosis  led  by  Oreste  Reale  

Example  2:  The  Quasi-­‐Biennial  Oscilla(on  

1990 1995 2000 2005 2010 2015

-40

-20

0

20

40

U (

ms

-1)

Mean Diff (ms-1): -1.01 1.98 Sdev Diff (ms-1): 4.15 8.26 10 hPa (104oE, 1oN)aSingaporeMERRAMERRA-2

1990 1995 2000 2005 2010 2015

-40

-30

-20

-10

0

10

20

U (

ms

-1)

Mean Diff (ms-1): 0.666 0.759 Sdev Diff (ms-1): 2.87 2.19 30 hPa (104oE, 1oN)b

1990 1995 2000 2005 2010 2015

-30

-20

-10

0

10

20

U (

ms

-1)

Mean Diff (ms-1): -1.55 0.158 Sdev Diff (ms-1): 2.68 1.74 50 hPa (104oE, 1oN)c

Singapore  (gray  shading)  MERRA  MERRA-­‐2  

MERRA-­‐2  fits  the  Singapore  observa(ons  bemer,  especially  during  the  westerly  phase  of  the  QBO.  

Work  led  by  Larry  Coy  

Zonal  wind  forcing  terms  in  MERRA  

0 1 2 3 4 5Years

100

10

1

Pre

ssu

re (

hP

a)

DUDT GWD (m s-1 month-1)

-8-8

4

4

4

12 12

-2

-2

0 0

0

0

00 0

2 2

a

0 1 2 3 4 5Years

100

10

1

Pre

ssu

re (

hP

a)

DUDT ANA (m s-1 month-1)

-8-8

4

4

4

12 12

0 00

0

0

0 02 2

b

0 1 2 3 4 5Years

100

10

1

Pre

ssu

re (

hP

a)

DUDT DYN (m s-1 month-1)

-8-8

4

4

4

12 12

-2 -2

0

00

0

0

0 0

2 2

c

0 1 2 3 4 5Years

100

10

1

Pre

ssu

re (

hP

a)

DUDT SUM (m s-1 month-1)

-8-8

4

4

4

12 12-2 -2

0 0

0

0

0 0

2

2

2

d

-36 -32 -28 -24 -20 -16 -12 -8 -4 4 8 12 16 20 24 28 32 36

U (m s-1)

Tropical  winds  in  MERRA  are  forced  in  roughly  equal  measure  by  resolved  dynamics,  gravity-­‐wave  drag,  and  analysis  increments    Excessive  reliance  on  analysis  to  force  these  winds  suggests  that  a  forcing  term  is  missing      Developments  of  the  GWD  scheme  (adding  an  extra  wave  spectrum  in  the  Tropics)  was  a  part  of  the  development  from  MERRA  to  MERRA-­‐2.        

Zonal  wind  forcing  terms  in  MERRA-­‐2  

0 1 2 3 4 5Years

100

10

1

Pre

ssu

re (

hP

a)

DUDT GWD (m s-1 month-1)

-8-8

4

4

4

12

12

-6 -6-4

-4

-4

-2

-2

-2

-2

0 0

0

0

0 0

2

2

2

2

2

4

4

4

6 6

a

0 1 2 3 4 5Years

100

10

1

Pre

ssu

re (

hP

a)

DUDT ANA (m s-1 month-1)

-8-8

4

4

4

12

12

0 0

0

0

0

0

0

0

0

b

0 1 2 3 4 5Years

100

10

1

Pre

ssu

re (

hP

a)

DUDT DYN (m s-1 month-1)

-8-8

4

4

4

12

12

-4 -4

-2 -2

-2 -20 0

0

0

0 0

22

2

2

2 2

2

2

4

4

c

0 1 2 3 4 5Years

100

10

1

Pre

ssu

re (

hP

a)

DUDT SUM (m s-1 month-1)

-8-8

4

4

4

12

12-2

-2

0 0

0

0

0 0

2

2

2

d

-36 -32 -28 -24 -20 -16 -12 -8 -4 4 8 12 16 20 24 28 32 36

U (m s-1)

The  impact  of  these  changes  to  GWD  is  high    The  physical  forcing  (GWD  driving)  now  dominates  the  QBO  wind  forcing  in  MERRA-­‐2      Analysis  increments  play  a  much  smaller  role  

Coy  et  al.  (2015)  

Example  3:  Aerosol  Analysis  in  MERRA-­‐2  

Diagnosis  led  by  Cynthia  Randles  

Ø  Unique  amongst  its  peers,  the  MERRA-­‐2  reanalysis  includes  an  aerosol  reanalysis  for  the  modern  satellite  era.  

Ø  Constrained  by  observed  aerosol  op(cal  depth  (AOD),  MERRA-­‐2  simulates  major  aerosol  events  (i.e.  volcanic  erup(ons)  as  well  as  the  temporal  and  spa(al  variability  of  major  aerosol  species.  

Sea  Salt   Dust  

Sulfate   Carbon  

Aircra3  observa6ons  of  ex6nc6on  ver6cal  profile  

Aircra3  observa6ons  of  AOD  R2  =  0.72  

Historical  Cruises  

[data  from  A.  Smirnov]  

Emerging  Applica(on  of  MERRA-­‐2  Aerosols  to  Climate  Studies  Du

st C

once

ntra

tion

[μg

m-3

]

0

40

80

120

160

200

240

J F M A M J J A S O N D

2006 Daily-Mean Dust Surface Concentration at Barbados

MERRA-2Prospero et al.

R2 = 0.692

240

High  correla(on  with  daily  surface  dust  at  Barbados  

Similar  long-­‐term  seasonal  cycle  of  surface  dust  

Saharan  Air  Layer  (SAL)  and  Aerosol-­‐Climate  Interac(ons  

[Figure  from  H.  Yu]  

Zonal  decay  in  AOD  from  African  source  to  Caribbean  matches  MODIS  C5  

AOD zonally-averaged between 0o

- 30o

N

-100 -80 -60 -40 -20 0

longitude

0.0

0.2

0.4

0.6

0.8

1.0

AO

D

MERRA-2 MODIS Aqua

Africa  

Carib

bean   [Figure  from  P.  Colarco]  

OMI  AAOD  

M2  AAOD  

OMI  –  M2  

Aerosol  absorp(on  effects  on  radia(on  and  climate  

[Figure  from  V.  Buchard]  

Summary  GEOS-­‐5  modeling  work  is  in(mately  linked  to  observa(ons:    •  Timescales  of  hours  to  decades,  includes  predic(on  of  weather  

and  seasonal-­‐to-­‐decadal  varia(ons    •  Data  assimila(on,  including  reanalysis,  is  a  core  element    Three  examples  of  model-­‐data  combina(ons,  with  a  focus  on  the  new  GEOS-­‐5  reanalysis,  called  MERRA-­‐2:    •  Observa(onally  derived  implementa(on  of  surface  drag  over  

oceans  –  beneficial  impacts  (when  care  is  taken)    •  Physical  forcing  of  the  QBO  by  gravity  wave  drag    •  Emerging  capabili(es  in  aerosol  assimila(on  –  direct  effects  

included  in  MERRA-­‐2