Scribd
Upload
56 views

What's The Difference Between FNL and GFS?

FNL and GFS are global weather forecast models produced by NOAA that use the same underlying data and model, but FNL incorporates about 10% more observational data than GFS, resulting in a more complete analysis. However, it is delayed by 60-90 minutes to allow for more data to be collected. GFS must start earlier to produce a forecast, while FNL is focused on accuracy over timeliness.

Uploaded by

Kholis Nur Cahyo
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
56 views

What's The Difference Between FNL and GFS?

FNL and GFS are global weather forecast models produced by NOAA that use the same underlying data and model, but FNL incorporates about 10% more observational data than GFS, resulting in a more complete analysis. However, it is delayed by 60-90 minutes to allow for more data to be collected. GFS must start earlier to produce a forecast, while FNL is focused on accuracy over timeliness.

Uploaded by

Kholis Nur Cahyo
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

What’s

 the  difference  between  FNL  and  GFS?  


 
FNL  and  GFS  are  related,  yet  different  products  from  the  same  data  assimilation  and  
forecast  system.    They  share  the  same  underlying  model  and  data  assimilation  
techniques.    They  contain  the  same  data  sources-­‐-­‐but  there  is  a  subtle  difference  in  
the  amount  of  "real"  data  assimilated  into  the  initial  conditions  for  GFS  and  FNL.      
 
It  takes  time  to  run  a  global  NWP  model,  even  with  NCEP's  vast  computational  
resources.    Thus,  they  need  to  start  GFS  early  enough  to  get  a  forecast  instead  of  a  
hindcast.      
 
FNL  is  the  final  analysis,  delayed  a  bit  from  GFS  so  that  they  can  include  all  of  the  
available  observational  data.    Typically,  FNL  ingests  about  10%  more  observations  
than  GFS.    Even  with  the  late  start,  it  is  still  available  in  time  so  that  the  6-­‐hour  FNL-­‐
based  forecast  can  be  used  as  the  background  field  for  the  next  GFS  data  
assimilation  cycle.      
 
For  instance,  if  you  want  balloon  data  from  00Z,  you  have  to  wait  for  the  balloon  to  
rise  through  the  atmosphere.    That  can  be  as  much  as  90  minutes  for  the  large  size  
balloons  capable  of  reaching  the  stratosphere.      Then  the  balloon  data  needs  to  be  
relayed  from  locales  around  the  world  to  NCEP  in  Maryland,  USA.      
 
If  you  want  satellite  data  for  00Z,  you  have  to  wait  for  the  satellite  to  go  over  a  
ground  station  so  it  can  download  the  data.    The  ground  station  then  relays  the  data  
via  land  or  undersea  cable  (if  it  has  one)  or  hopscotches  it  through  another  
(communication)  satellite.      
 
A  table  at  the  link  below  shows  the  actual  initiation  time  for  each  analysis/forecast  
cycle  initiated  (30  day  running  average).        
http://www.nco.ncep.noaa.gov/pmb/nwprod/prodstat/index.html#TARGET      
 
The  first  picture  is  an  image  of  the  top  of  the  current  status  page.      
 
 
 
The  00  UTC  GFS  cycle  looks  like  this;  scroll  down  to  see  the  06,  12,  18  UTC  statuses.  

 
 
Studying  these  tables  gives  you  an  idea  of  the  tempo  and  complexity  of  numerical  
weather  prediction.      
 
The  GFS  analysis  for  00Z  doesn't  finish  until  ~01:22Z.  NCEP  waits  for  more  data  to  
roll  in,  and  then  creates  the  FNL  analysis  with  the  more  complete  data  set.  NCEP  
then  runs  the  forecast  out  6  hours  using  FNL  as  the  starting  point.    That  6-­‐hour  
forecast  is  used  as  the  background  field  for  the  next  GFS  analysis  at  06Z.    The  table  
uses  these  abbreviations:      
NAM  =  North  American  Mesoscale        
GFS  =  Global  Forecast  System        
AVN  =  Aviation        
 
If  GFS  is  initialized  at  2014-­‐08-­‐01_00,  then  it  uses  fnl_2014073118_06  as  a  
background  field  before  data  assimilation.    That  means  the  FNL  analysis  cycle  from  
2014-­‐07-­‐31  at  18Z,  but  forecast  out  for  6  hours.    That  makes  the  file  valid  for  2014-­‐
08-­‐01  at  00Z.      
 
FNL  is  not  on  the  table  because  it  is  not  disseminated  as  an  operational  product  (in  
the  sense  that  it  is  available  24x7,  in  real-­‐time,  without  data  drops).            
 
In  summary,  the  FNL  analysis  incorporates  ~10%  more  observational  data  than  the  
GFS  analysis  at  the  cost  of  a  60-­‐90  minutes  delay.      
 
As  an  archive,  the  RDA  is  interested  in  offering  the  most  realistic  atmospheric  
analysis,  not  the  timeliest  one.      Users  shouldn't  come  to  an  archive  for  timely  data-­‐-­‐
they  should  come  for  the  most  complete  and  accurate  data  possible.  We  only  archive  
the  analysis  FNL  file-­‐-­‐those  that  end  in  _00.      The  forecast  ones  end  in  _HH  for  the  
number  of  hours  from  model  initiation.        "Real-­‐time"  users  can  get  the  FNL  analysis  
grids  directly  from  NCEP's  NOMADS  server,  http://nomads.ncep.noaa.gov/.  
 
To  learn  more,  read  http://rda.ucar.edu/datasets/ds083.2/docs/Analysis.pdf  
 
Grace  Peng,  PhD  
Atmospheric  &  Geoscience  Research  Data  Archive  
Computational  &  Information  Systems  Laboratory  
National  Center  for  Atmospheric  Research  
303-­‐497-­‐1218  
5  December  2014  
 
 
 

You might also like