40
My Agenda for CFS Diagnostics [email protected] Ancient Chinese proverb: “ Even a 9-month forecast begins with a single time step.” --Hua-Lu Pan

My Agenda for CFS Diagnostics [email protected] Ancient Chinese proverb: “ Even a 9-month forecast begins with a single time step.” --Hua-Lu Pan

  • View
    221

  • Download
    3

Embed Size (px)

Citation preview

My Agenda for CFS Diagnostics

[email protected]

Ancient Chinese proverb:

“ Even a 9-month forecast begins with a single time step.”

--Hua-Lu Pan

IMHO

--diagnostics of hind-casts and AMIP/CMIP integrations complicated

--no atmospheric reanalysis consistent with CFS

--significant uncertainties in air-sea fluxes

--single global forecast model for weather and seasonal forecasts

--operational weather forecasts give larger sample of performance of atmospheric model

--should examine carefully short-range CFS forecasts

--more insight into model problems from 1-day forecasts than from seasonal forecasts or AMIP/CMIP integrations

Hypothesis--

Over last 10 years considerable effort at EMC in improving tropical analyses and short-range forecasts has produced good atmospheric model for seasonal forecasts of equatorial Pacific SST.

Better hurricanes Better ENSO

Moral--

Don’t separate weather and climate.

“Error” in transient eddy kinetic energy in 64 level AMIP run of GFS

Sep. 01-Mar.02

1979-2001

Difference in analyzed transient eddy kinetic energy Sep.01 –Mar.02

GDAS minus NCEP1

NCEP1—T62, 28 levels

NESDIS temperature retrievals

GDAS—T170, 42 levels

Radiances used

Transient eddies in AMIP runs cannot be verified in the Southern Hemisphere;

error in AMIP transient eddies appears to be less than error in NCEP1 reanalysis.

Verifying AMIP runs of the current GFS model against NCEP1 reanalysis can be inappropriate, even for atmospheric fields.

Need to compare CDAS to GDAS to see if differences between CDAS and GDAS are insignificant compared to differences between AMIP and CDAS.

Need periodic reanalyses with the operational CFS for consistent verification of seasonal forecasts.

GODAS (MOM V.3)

Forced by wind stress, heat flux, precipitation-evaporation

SST is relaxed to weekly NCEP SST analysis

surface salinity is relaxed to Levitus monthly SSS climatology.

Wind stress is thought to have most impact

OGCMMOM

Global v.3

DataAssimilation

3D VAR

Observations:XBTsTAO

P-FloatsAltimetry

AnalyzedFields:

TemperatureSalinity

Ocean Data Assimilation System (ODAS)

Surface StressHeat Flux

P-EFrom ?

--NCEP-2 reanalysis 1979-present (CDAS-2)T62, 28 levels Used in hind-castsnot consistent with CFS

What are best air-sea fluxes to initialize ocean assimilation for

real-time forecasts?

--GDAS T254, 64 levels

better data assimilation and atmospheric model than NCEP-2 better fluxes??

Not used in hind-castsMore consistent with CFS than NCEP2

NCEP2 too many easterly waves (Hodges et al., J Clim)

Normalized RMS difference of monthly mean stress over 3 years

Normalized bias in 3-year mean stress magnitude

GDAS vs.

CDAS

Nino 3.4

5S-5N

190-240E

Equatorial east Pacific

Correlation monthly zonal surface stress anomalies 1979-2001

FSU-CDAS2

FSU-ERA40

CDAS2-ERA40

29S-29N 5S-5N

FSU-NCEP2 .63 .55

FSU-ERA40 .68 .65

NCEP2-ERA40 .85 .75

Correlation of monthly anomalies in zonal wind

stress 122-290E 1979-2001

29S-29N 5S-5N

FSU-NCEP2 .59 .58

FSU-GDAS .63 .67

NCEP2-GDAS .87 .77

Correlation of monthly anomalies in zonal wind

stress 122-290E July 2001-Dec. 2003

Bias 0.76

RMS 19.7

SD 19.6

Bias 19.16

RMS 20.8

SD 10.9

2002-2003 minus 1985-1990

  SOC 80-93

SRB183-91

NCEP1 80-93

NCEP2 80-93

ERA1581-92

CDAS2Aug02-Jul03

operAug02-Jul03

LH 90   92 103 103 113 109

SH 7   11 5 10 6 6

NSW 176 173 165 167 160 169 192

NLW 49 42 57 51 50 51 67

NHF 30   6 7 -4 -1 11

Global Mean Ocean Heat Budget

Precipitation

Annual mean

CMAP

GDAS

CDAS2

CONCLUSIONS--GDAS, ERA40 surface stresses agree more with independent estimates than NCEP2, implying progress

--disagreement between different estimates in equatorial Pacific implies substantial uncertainty in surface stress

--NCEP2 too many easterly waves in Pacific

--surface heat flux global balance in NCEP2, not GDAS

--GDAS too much sfc NSW; clouds, moisture need work

--GDAS better patterns of sfc NSW, NHF

--GDAS better precipitation pattern than NCEP2

New CFS will use fluxes from NCEP-2 reanalysis to force ocean data assimilation, for consistency with hind-casts

Future CFS will conduct reanalyses with new CFS models for consistency of system as well as consistency with hind-casts. (If $$)

New CFS every 3-5 years.

New global reanalyses 1979 to present every 3-5 years in support of seasonal forecasting.

Climate drift in SST in 23 year coupled runs

Climate drift with 28 levels

Climate drift with 64 levelsMuch less drift along equator with 64 levels

Surface stress Pacific

22 years FSU obs estimates

23 year coupled runs

FSU

28 levels

64levels

Ekman pumping

28 level run has too much upwelling along equator in Pacific

East Pacific

1) Increased Ekman pumping and more upwelling at equator in Pacific with 28 than with 64 levels, probably associated with stronger Pacific ITCZ with 28 levels; doesn’t explain warmer tropical SST in Atlantic

2) Lack of marine stratus with 64 levels compared to 28 levels; more short wave in eastern oceans, possibly warming SST which are advected to equator.