Recent and planed NCEP climate modeling activities

1
Recent and planed NCEP climate modeling activities

• Hua-Lu Pan
• EMC/NCEP

2

• EMC is considering the following upgrades to the
  operational CFS
• Reducing the 8-lag in the initial condition to a
  1-day lag.
• Introducing 2 new members at higher resolution
  (T126L64) out to 60 days only. These 2 members
  would initiate from the same initial condition as
  the operational two T62L64 members.
• Both these upgrades would aim to improve upon the
  week 3-6 / monthly forecast leads.

3
NAO PNA SKILL CFS Operational T62L64
Forecasts MAR2005 FEB2007 2 IC at Day 1-7 of
each of the 24 months Forecast lead times 1-45
days
4
Negative skewness in observations
5
(No Transcript)
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
Daily skill scoresAveraged over 14 forecasts
per monthafter forecast lead time of 8 days
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
Studies to prepare for a monthly forecast suite
(week3 to week6)

• Four-times daily T126 version of the CFS,
• Runs are made to 60 days in length,
• Daily runs of the hindcasts,
• The test period is 2000-2004.
• The period of the runs are 7 May 15 July and
  7 Nov 15 Jan.

15
Daily
16
Month-1 Correlation Scores NAO 33.0 PNA 21.3
17
Impact of Coupling
18
Pattern correlation for 2000-2004
Summer
0.6
0.4
Persistence
Winter
0.6
0.4
Persistence
19
Pattern Correlation for initialization dates from
May to June 2002
The Predictability Barrier
June 6th-9th
June 6th-9th
June 6th-9th
6-9 June MJO maximum activity crosses the
Maritime Continent
20
March 2005 Feb 2007
Week 1
Week 2
Week 3
Week 4
Monthly
Longitude
21
March 2005 Feb 2007
0.55 mean ac of daily forecasts in week 1 range
Week 1
Week 2
Week 3
Week 4
Monthly
Initial Month
22
Design of the 30-year NCEP
CFSRR T382L64 Global Reanalysis and T126L64
Seasonal Reforecast Project(1979-2008)
Suru Saha and Hua-Lu Pan, EMC/NCEP With Input
from Stephen Lord, Mark Iredell, Shrinivas
Moorthi, David Behringer, Ken Mitchell, Bob
Kistler, Huug van den Dool and others
23
NCEP CFSRR
SLIDE - 1

• 4 Simultaneous Streams
• Jan 1978 Sep 1987 9 years
• Oct 1985 Sep 1994 9 years
• Oct 1992 Sep 2001 9 years
• Oct 1999 Sep 2008 9 years
• 2-year overlap for ocean and land spin ups
• Total of 36 years of Reanalysis

24
NCEP CFSRR
SLIDE - 3
ONE DAY OF REANALYSIS

• Atmospheric T382L64 (GSI) Analysis at 0,6,12 and
  18Z
• Ocean and Sea Ice Analysis (GODAS) at 0,6,12 and
  18Z
• From each of the 4 cycles, a 9-hour coupled
  guess forecast (GFS at T382L64) is made with
  hourly coupling to the ocean (MOM4 at 1/4o
  equatorial, 1/2o global)
• Land (GLDAS) Analysis with Noah at 0Z
• Coupled 2-day forecast from initial conditions
  from every cycle, or coupled 5-day forecast from
  every 0Z cycle (starting every year from 1 Jan
  00Z) is made with the T382L64 GFS with hourly
  coupling to the ocean (MOM4 at 1/4o equatorial,
  1/2o global) for sanity check.

25
NCEP CFSRR
SLIDE - 4
CFS REFORECASTS
1 Jan 0Z
2 Jan 6Z
3 Jan 12Z
4 Jan 18Z
6 Jan 0Z
1 year T126L64 coupled forecast ( GFS MOM4
Noah )
Coupled one-year forecast from initial conditions
30 hours apart is made for 2 initial months
(April and October) with the T126L64 GFS with
hourly coupling to the ocean (MOM4 at 1/4o
equatorial, 1/2o global). Total number of
forecasts 28 x 2 x 30 1680 For each cycle,
there will be approximately 7 members per month,
with a total of 210 members over a 30-year
period. This ensures stable calibration for
forecasts originating from each cycle, for a
given initial month
26
NCEP CFSRR
SLIDE - 5
CFS REFORECASTS
Example Skill Estimation for the Initial Month
of February In real time operational prediction,
the latest members would be used. For instance,
the earliest forecasts release for CPC would be
the 15th of the month. Therefore, all members up
to the 9th of the month would be used since
operations would run in near real time. For
skill estimation, the following 16 members would
be used for February starts (going backwards)
Feb 8 18z, Feb 7 12z, Feb 6 06z, Feb 5 00z, Feb
3 18z Jan 21 00z This includes 4 days of 4
different cycles (Jan 21 0z Feb 8 18z),
smoothly separated by 30 hours. This is in
contrast to the present system, where we use 15
members from Jan 9 00z Feb 3 00z, in 5-day
chunks.