How Does NCEP/CPC Make Operational Monthly and Seasonal Forecasts?

1

How Does NCEP/CPC Make Operational Monthly and Seasonal

Forecasts?

Huug van den Dool (CPC)

CPC, June 23, 2011/ Oct 2011/ Feb 15, 2012/ UoMDMay,2,2012/ Aug2012/ Dec,12,2012/UoMDApril24,2013/May22,2013,/Nov20,2013/April,23,2014/

2

Assorted Underlying Issues

• Which tools are used…

• How do these tools work?

• How are tools combined???

• Dynamical vs Empirical Tools

• Skill of tools and OFFICIAL

• How easily can a new tool be included?

• US, yes, but occasional global perspective

• Physical attributions

3

Menu of CPC predictions:

• 6-10 day (daily)• Week 2 (daily)• Monthly (monthly + update)• Seasonal (monthly)• Other (hazards, drought monitor, drought outlook,

MJO, UV-index, degree days, POE, SST) (some are ‘briefings’)

• Operational forecasts (‘OFFICIAL’) and informal forecast tools (too many to list)

• http://www.cpc.ncep.noaa.gov/products/predictions/90day/tools/briefing/index.pri.html

4

EXAMPLEPUBLICLY

ISSUED

“OFFICIAL”FORECAST

5

6

7From an internal CPC Briefing package

8

EMP EMP EMP EMP

EMPDYN

DYN

CONCON

N/A

99

SMLR CCA OCN

LAN

LFQ

(15 CASES: 1950, 54, 55, 56, 64, 68, 71, 74, 75, 76, 85, 89, 99, 00, 08)

OLD-OTLK

CFSV1

ECPIRI

ECA CON

10

Element US-T US-P SST US-soil moisture

Method:CCA X X XOCN X X CFS X X X XSMLR X XECCA X XConsolidation X X X

Constr Analog X X X XMarkov XENSO Composite X XOther (GCM) models (IRI, ECHAM, NCAR, N(I)MME):

X X

CCA = Canonical Correlation AnalysisOCN = Optimal Climate NormalsCFS = Climate Forecast System (Coupled Ocean-Atmosphere Model)SMLR = Stepwise Multiple Linear RegressionCON = Consolidation

11

Long Lead Predictions of US SurfaceTemperature using Canonical CorrelationAnalysis. Barnston(J.Climate, 1994, 1513)

Predictor - Predictand Configuration

Predictors Predictand

* Near-global SSTA

* N.H. 700mb Z * US sfc T

* US sfc T

four predictor “stacked” fields one predictandperiod

4X652=2608 predictors 102 locations

Data Period 1955 - last month

12

13

14

About OCN. Two contrasting views:- Climate = average weather in the past- Climate is the ‘expectation’ of the future

30 year WMO normals: 1961-1990; 1971-2000; 1981-2010 etc

OCN = Optimal Climate Normals: Last K year average. All seasons/locations pooled: K=10 is optimal (for US T).

Forecast for Jan 2015 (K=10) = (Jan05+Jan06+... Jan14)/10. – WMO-normal

plus a skill evaluation for some 50+ years.

Why does OCN work?1) climate is not constant (K would be infinity for constant climate)2) recent averages are better3) somewhat shorter averages are better (for T)see Huang et al 1996. J.Climate. 9, 809-817.

15

OCN has become the bearer of most of the skill, see also EOCN method

(Peng et al), or other alternatives of projecting normals forward.

16

17

[email protected]

GHCN-CAMS

FAN

2008

18

[email protected]

Preview of 2010s, 4 years only

NCEP’s Climate Forecast System, now called CFS v2

• MRFb9x, CMP12/14, 1995 onward (Leetmaa, Ji etc). Tropical Pacific only.

• SFM 2000 onward (Kanamitsu et al• CFSv1, Aug 2004, Saha et al 2006. Almost global

ocean• CFSR, Saha et al 2010• CFSv2, March 2011. Global ocean, interactive

sea-ice, increases in CO2. Saha et al 2014.19

NCEP’s Climate Forecast System, now called CFS v2

20

<--Out of date diagram.Still instructive

21

22

23

Major Verification Issues

• ‘a-priori’ verification (used to be rare)

• After the fact (fairly normal and traditional)

24

Source Peitao Peng

After the fact…..

25

(Seasonal) Forecasts are useless unless accompanied by a reliable a-

priori skill estimate.

Solution: develop a 50+ year track record for each tool. 1950-present.

(Admittedly we need 5000 years)

26

Consolidation

27

--------- OUT TO 1.5 YEARS -------

28

OFFicial Forecast(element, lead, location, initial month) =

a * A + b * B + c * C +…

Honest hindcast required 1950-present. Covariance (A,B), (A,C), (B,C), and

(A, obs), (B, obs), (C, obs) allows solution for a,

b, c (element, lead, location, initial month)

29

CFS v1 skill 1982-2003

30

Fig.7.6: The skill (ACX100) of forecasting NINO34 SST by the CA method for the period 1956-2005. The plot has the target season in the horizontal and the lead in the vertical. Example: NINO34 in rolling seasons

2 and 3 (JFM and FMA) are predicted slightly better than 0.7 at lead 8 months. An 8 month lead JFM forecast is made at the end of April of the previous year. A 1-2-1 smoothing was applied in the vertical to

reduce noise.CA skill 1956-2005

31

M. Peña Mendez and H. van den Dool, 2008: Consolidation of Multi-Method Forecasts at CPC. J. Climate, 21, 6521–6538.

Unger, D., H. van den Dool, E. O’Lenic and D. Collins, 2009: Ensemble Regression.Monthly Weather Review, 137, 2365-2379.

(1) CTB, (2) why do we need ‘consolidation’?

32

33

(Delsole 2007)

34

3CVRE

SEC

SEC and CV

35

36

37

38

39

40

41

42

See also:

O’Lenic, E.A., D.A. Unger, M.S. Halpert, and K.S. Pelman, 2008: Developments in Operational Long-Range Prediction at CPC. Wea. Forecasting, 23, 496–515.

43

Empirical tools can be comprehensive! (Thanks to

reanalysis, among other things).

And very economical.Constructed Analogue(next 2 slides)

• Given an Initial Condition, SSTIC (s, t0) at time t0 . We express SSTIC (s, t0) as a linear combination of all fields in the historical library, i.e.

2012 or 2013• SSTIC (s, t0) ~= SSTCA(s) = Σ α(t) SST(s,t) (1)

t=1956 or 1957 (CA=constructed Analogue)

• The determination of the weights α(t) is non-trivial, but except for some pathological cases, a set of (57) weights α(t) can always be found so as to satisfy the left hand side of (1), for any SSTIC , to within a tolerance ε.

• Equation (1) is purely diagnostic. We now submit that given the initial condition we can make a forecast with some skill by

2012 or 2013• XF (s, t0+Δt) = Σ α(t) X(s, t +Δt) (2)

t=1956 or 1957Where X is any variable (soil moisture, temperature, precipitation)• The calculation for (2) is trivial, the underlying

assumptions are not. We ‘persist’ the weights α(t) resulting from (1) and linearly combine the X(s,t+Δt) so as to arrive at a forecast to which XIC (s, t0) will evolve over Δt.

Year Wgt Year Wgt Year Wgt Year Wgt Year Wgt Year Weigt

1956 -5 1966 -10 1976 -4 1986 5 1996 2 2006 2

1957 12 1967 0 1977 0 1987 -9 1997 9 2007 2

1958 3 1968 1 1978 -3 1988 -10 1998 -11 2008 2

1959 13 1969 -6 1979 -3 1989 0 1999 -2 2009 11

1960 -7 1970 -4 1980 8 1990 5 2000 -17 2010 6

1961 -2 1971 2 1981 -7 1991 14 2001 3 2011 -1

1962 5 1972 4 1982 -12 1992 -3 2002 -2 2012 12

1963 5 1973 10 1983 -7 1993 -4 2003 20 2013 7

1964 -8 1974 6 1984 3 1994 -7 2004 -1 2014 NA

1965 -9 1975 -2 1985 2 1995 -1 2005 7 Xx

CA-weights inMarch 2014

47

48

SST Z500

PrecipT2m

CA

49

SST Z500

PrecipT2m

CFS

Source: Wanqiu Wang

Physical attributions of Forecast Skill

• Global SST, mainly ENSO. Tele-connections needed.

• Trends, mainly (??) global change

• Distribution of soil moisture anomalies

50

Website for display of NMME&IMME

NMME=National Multi-Model EnsembleIMME=International Multi-Model Ensemble

• http://origin.cpc.ncep.noaa.gov/products/NMME/

Please attend

• Friday 2pm June 14

• Tuesday 1:30pm June 18

Two meetings to Discuss the Seasonal Forecast.

52