Assessing the representation of north-‐west Australian rainfall in GCMs

16
Assessing the representation of northwest Australian rainfall in GCMs Duncan Ackerley Gareth Berry, Christian Jakob and Michael Reeder Monash University [email protected]

Transcript of Assessing the representation of north-‐west Australian rainfall in GCMs

Assessing  the  representation  of  north-­‐west  Australian  rainfall  

in  GCMs    

Duncan  Ackerley  

Gareth  Berry,  Christian  Jakob  and  Michael  Reeder  

Monash  University  

 

[email protected]  

Why  do  we  care  about  north-­‐west  Australian  rainfall?    

•  Figure:  1950  –  2008  linear  trend  in  DJF  rainfall  (mm  

yr-­‐1).  

•  Only  positive  trend  

shown.      

•  Strong  increase.  •  Increased  water  supply  –  

resource?  

[email protected]  

•  Data  from  Bureau  of  Meteorology.    

•  Several  hypotheses  as  to  the  cause:  

–  Aerosol,  Tropical  Atlantic  SST,  Indian  Ocean  SST  and  ozone  depletion.    

•  Different  models  have  opposing  signs  for  the  rainfall  change  associated  

with  ozone  depletion.    

•  AR4:  8-­‐13  models  out  of  21  predicted  increased  rainfall  for  north-­‐west  

Australia:  

–  Little  consensus?  

–  Precipitation  change  is  small?  

–  Combination  of  the  two?  

•  It  therefore  seemed  sensible  to  go  back  and  evaluate  whether  any  model  

could  actually  represent  the  physical  mechanisms  correctly  before  we  

could  make  any  statements  about  the  cause  of  the  trend.  

[email protected]  

How?    

•  Start  simple:  Assess  our  own  model  in  Australia.    

•  Use  ACCESS1.3  general  circulation  model.    

•  Simulate  December-­‐January-­‐February  (DJF)  1979/80  –  2007/08.  

•  Forced  using  AMIP  boundary  conditions.    

 

•  Ask  the  question  –  can  ACCESS  represent  the  physical  processes  that  lead  to  rainfall  in  north-­‐west  

Australia?  

•  Use  ACCESS  as  a  test  case  to  evaluate  the  important  processes  that  lead  to  rainfall.    

–  Seasonal  mean  climatology  (not  shown).  

–  Diurnal  cycle.  

–  Air  flow  and  moisture  sources.  

•  These  analyses  can  then  be  applied  to  AMIP  simulations  of  other  models.  

[email protected]  

Diurnal  cycle  of  the  heat  low  circulation.    

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

1010.0

1012.0

1014.0

1016.0

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

(a) ERA-Int 08 AWST

1010

10

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S1006.0

1008.01010.01012.0

1014.01016.0

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

(b) ERA-Int 14 AWST

1010

10

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

1008.01010.0

1012.01014.01016.0

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

(c) ERA-Int 20 AWST

1010

10

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

1008.0

1010.01012.0

1014.0

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

(d) ERA-Int 02 AWST

1010

10

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

1008.0

1010.01012.0

1014.01016.0

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

(e) ACCESS 1.3 08 AWST

1010

10

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

1004.0

1006.0

1008

.0

1010.01012.01014.01016.0

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

(f) ACCESS 1.3 14 AWST

1010

10

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

1006.0

1008.0

1010.01012.01014.01016.0

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

(g) ACCESS 1.3 20 AWST

1010

10

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

1006.0

1008

.0

1010.01012.01014.0

1016.0

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

(h) ACCESS 1.3 02 AWST

1010

10

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

0.0 1.0 2.0 3.0 4.0 5.0 6.0Wind speed (m s-1)

•  6-­‐hourly  MSLP  and  925  hPa  wind  

composites  for  all  DJFs  over  AMIP  

period.    

•  AWST  –  Australian  Western  Standard  

Time  (UTC  +  8).    

•  Shading  indicates  regions  of  low  wind  

speed.    

 

•  Simple  proxy  for  identifying  the  centre  

of  the  heat  low  circulation.  

   

•  ERA-­‐Interim  and  ACCESS1.3  agree  well  

during  the  day.    

•  Disagreement  overnight.    

•  Heat  low  circulation  displaced  in  

ACCESS1.3.    

•  E  to  NE  flow  tendency  in  ERA-­‐Interim.  

•  E  to  SE  flow  tendency  in  ACCESS1.3.      

[email protected]  

6-­‐hourly  ageostrophic  flow  and  convergence.    

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

55

5

(a) ERA-Interim: 08 AWST

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

55

5

(b) ERA-Interim: 14 AWST

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

55

5

(c) ERA-Interim: 20 AWST

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

55

5

(d) ERA-Interim: 02 AWST

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

55

5

(e) ACCESS1.3: 08 AWST

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

55

5

(f) ACCESS1.3: 14 AWST

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

55

5

(g) ACCESS1.3: 20 AWST

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

55

5

(h) ACCESS1.3: 02 AWST

110E 120E 130E 140E 150E40S

35S

30S

25S

20S

15S

10S

•  Ageostrophic  flow  vectors  at  925  hPa  .    

•  Convergence:  shaded.    

•  Convergence  occurs  near  the  coast  

during  the  day.    

•  Nocturnal  convergence  occurs  inland.    

•  Similar  patterns  in  ERA-­‐Interim  and  

ACCESS1.3.    

•  However,  inland  ageostrophic  flow  

penetrates  further  south  and  west  into  

north-­‐west  Australia  (ERA).    

•  HOW  DOES  THIS  INFLUENCE  

RAINFALL?  

[email protected]  

6-­‐hourly  rainfall  as  a  percentage  of  the  daily  mean  total  

•  TRMM  3B42  data:  1998/99  –  

2007/08.  

•  ACCESS1.3:  1979/80  –  2007/08.  

•  Rainfall  occurs  too  early  in  much  

of  eastern  Australia.    

•  TRMM  and  ACCESS1.3  compare  

well  in  the  afternoon.    

 

•  Nocturnal  rainfall  maximum  

shifted  south  and  west  in  

ACCESS1.3.    

•  No  rainfall  at  02-­‐08  AWST  anyway,  

so  good  agreement  then.    

[email protected]  

Not  a  daily  occurrence…  

•  The  diurnal  cycle  dominates  in  the  

composites.  

•  However,  there  are  many  days  without  

rainfall.    

•  Therefore  the  diurnal  cycle  must  not  be  the  

only  factor  governing  rainfall.      

[email protected]  

Rainfall  initiation  

•  Perhaps  the  difference  comes  from  the  circulation?    

 

•  ERA-­‐Interim  925  hPa  geopotential  height  and  wind  

composites.    

•  Asked  the  question:  What  does  the  geopotential  

height  and  925  hPa  air  flow  look  like  on  dry  and  wet  

days  at  these  places?  

 

•  Top:  Wet  days  (when  rainfall  was  initiated)  at  

pluviograph  sites  (dots).  

•  Middle:  Dry  days.  

•  Bottom:  Differences.    

 

•  Enhanced  westerly  at  coast.  

•  Enhanced  north-­‐easterly  inland.    

[email protected]  

Image  taken  from  Figure  6  in:    Berry,  G.,  M.  J.  Reeder  and  C.  Jakob,  2011:  Physical  mechanisms  regulating  summertime  rainfall  over  northwestern  Australia.  Journal  of  Climate,  24,  3705-­‐3717.  ©  American  Meteorological  Society.    Used  with  permission.  

ACCESS1.3  rainfall  initiation  

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S(a) MP wet

700.0

720.0

740.0

760.0

780.0

800.0

8

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S(b) MP dry

700.0

720.0

740.0

760.0

780.0

800.0

8

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S(c) MP wet-dry

-13.0

-11.0

-9.0

-7.0

-5.0

-3.0-2.0

2.0

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S

4

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S(d) RF wet

700.0

720.0

740.0

760.0

780.0

800.0

8

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S(e) RF dry

700.0

720.0

740.0

760.0

780.0

800.0

8

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S(f) RF wet-dry

-13.0

-11.0

-9.0

-7.0

-5.0

-3.0-2.0

2.0 6.010.0

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S

4

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S(g) NP wet

700.0

720.0

740.0

760.0

780.0

800.0

8

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S(h) NP dry

700.0

720.0

740.0

760.0

780.0

800.0

8

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S(i) NP wet-dry

-13.0

-11.0

-9.0

-7.0

-5.0

-3.0-2.0

2.0

6.0

110E 120E 130E 140E 150E 160E30S

25S

20S

15S

10S

5S

4

•  Corresponding  composites  for  

ACCESS1.3.    

•  Weakened  westerlies  at  

the  coast  –  incorrect.    

•  North-­‐easterlies  inland  

–  correct.    

[email protected]  

Conclusions  (for  ACCESS1.3)  

•  The  diurnal  cycle  of  the  circulation  corresponds  well  between  ACCESS1.3  and  reanalyses;  however,  there  

are  some  spatial  differences.  

•  The  diurnal  cycle  of  rainfall  in  ACCESS  corresponds  well  with  observations  at  certain  times,  again  there  

are  spatial  differences  between  the  observations  and  model.  

 

•  North-­‐easterlies  responsible  for  bringing  moisture  inland  –  well  represented.    

•  Coastal  rainfall  initiation  is  poor  in  the  model.    

 

•  All  of  these  features  discussed  in  a  paper  submitted  to  Journal  of  Climate  –  speak  to  me  for  details.    

•  These  processes  in  particular  must  be  analyzed  in  models  before  we  can  even  start  to  look  at  forcings  –  

diurnal  cycle  and  day-­‐to-­‐day  initiation  processes.  

•  Method  described  above  can  be  applied  to  CMIP5  models  where  the  relevant  data  exist.        

•  So  what  about  other  models?  

[email protected]  

CMIP5  

•  Some  early  analysis  on  the  CMIP5  models.    

•  Similar  to  the  figures  above:  

–  6-­‐hourly  TRMM  accumulation  (1998-­‐2007)  

– Corresponding  figures  for  21  models  

– Data  taken  from  AMIP  simulations  (1979-­‐2008)  

–  Look  at  the  afternoon  and  nocturnal  cases.    

•  Interesting  similarities  and  differences!  

[email protected]  

1400  AWST  –  2000  AWST  

[email protected]  

2000  AWST  –  0200  AWST  

[email protected]  

Final  conclusions  •  Afternoon  rainfall:  

–  No  rainfall  in  some  models  –  all  rained  out  in  morning  (not  shown)  possibly  due  to  early  convection  scheme  

trigger.  

–  Approximately  8  models  capture  the  pattern  and  relative  amounts  quite  well.    

 

•  Nocturnal  rainfall:  

–  Difficult  test  for  the  models,  which  two  seem  to  pass.    

–  Some  models  (ACCESS  for  example)  have  a  nocturnal  peak  but  it  is  in  the  wrong  place,  which  is  still  a  good  

result!    

•  Next  step  is  to  go  back  and  identify  why  these  models  do  this  in  a  similar  way  to  ACCESS  (where  data  exist):  

–  Circulation  

–  Heat  low  

–  Moisture  sources  

•  Try  to  pick  the  best  one(s)  to  understand  what’s  forcing  the  increase  in  rainfall.    

•  TAKE  HOME  MESSAGE:  ASSESS  ALL  OF  THE  ABOVE  BEFORE  USING  A  MODEL  TO  EXPLAIN  THE  TREND!  

[email protected]  

0800  AWST  –  1400  AWST  

[email protected]