Decadal prediction

Decadal prediction

Scientific interest Meetings and Workshops Organizational aspects WGSIP role and activities

WGCM/WGSIP and Decadal Predictability

International scientific interest

CMIP5 Experiment Design

“Long-Term”(century & longer)

TIER 1

TIER 2

CORE

“realistic”

diagnostic

“Near-Term”(decadal)

(initialized ocean state)

prediction & predictability

CORE

TIER 1

Active meeting and workshop schedule

OceanObs09 (Venice, Sept 09) 8th Workshop on Decadal Climate Variability (Maryland,

Oct 09) Earth-System Initialization for Decadal Prediction (deBilt,

Nov 09) Predicting Climate of the Coming Decades (Miami, Jan 10) WGSIP-13 (Buenos Aires, July 10) Conference on Decadal Predictability (Triest, Aug, 10) Workshop on Decadal Variability, Predictability and

Predictions: understanding the role of the oceans (NCAR, Sept 10)

WGCM-14 (Exeter, Oct 10) Seasonal to Multi-decadal Predictability of the Polar

Climate (Bergen, Oct 10) IPCC 1st LA Meeting (Kunming, Nov 10)

OceanObs09 Decadal White Papers

Decadal prediction

Scientific interest Meetings and Workshops Organizational aspects

focus survey panel IPCC

WGSIP role and activities

PCMDI website

Decadal prediction focus? current CLIVAR site is really very good but perhaps “focus” on

decadal aspect? links from/to WGSIP, WGCM, WGOMD plus PCMDI to a CLIVAR

“decadal prediction” page? currently under cross-cuts lacks a “name”

what’s in a name? “near-term” (decadal) hindcast/prediction simulations (!) WCRP decadal prediction CMIP5 near term experiments ….

what should be in the name? prediction or forecast (not simulation or experiment) timescale (near term climate, decadal, …) activity (project, intercomparison, …) sponsor/identifier (WCRP, CLIVAR, CMIP5, …)

something like CMIP5 Decadal Prediction Project (CDPP) or CMIP5 Decadal Forecasting Project (CDFP) ….

Does WGSIP want to make recommendations in this area (i.e. focus and name) ?

Organizational aspects of decadal prediction: survey

SIP community natural participants Experience/ability in most needed

aspects IPCC-class models initialization, ensemble generation retrospective forecast methodology forecast combination, calibration, skill

measures … (external forcing)

WGSIP-12 Meeting Report

(include them?)(CMIP)

• should we survey both• other questions or or keep it simple• suitable wrt CHFP timeline

Results of draft CHFP survey

CHFP action items?

data considerations critical CIMA role email for test data timeline

what of data not sent to CIMA EMSEMBLES CCCma, IRI, …

role for BADC? mirror CIMA actively collect data from other above sites

Results of draft CDFP survey

Follow on activities …• complete survey• representative at WGCM meeting (Exeter, Oct.)• …

WGSIP-12 Meeting Report

CMIP5-WGSIP-WGCM panel

At the Paris WGCM meeting a CMIP5-decadal Panel was struck consisting of Tim Stockdale and/or GJB (WGSIP), M. Latif, R. Stouffer (CMIP), G. Meehl (WGCM)

I have included K. Taylor in any interactions such as they are

So far, the actions of the Panel have consisted in some discussion of the treatment of volcano forcing for decadal prediction

resulted in a useful clarification no other burning questions have arisen so far

Presume WGSIP is keen to entrain the seasonal/interannual prediction (SIP) community into CMIP5-decadal

very natural extension from SIP to seasonal/decadal (S2D) prediction has useful experience

Item 17: “panel”

Email to CMIP5-WGSIP-WGCM panel

Suggesting potential actions which could be discussed at the upcoming WGSIP meeting in July:

to circulate an email to members of the community urging participation in CMIP5-decadal and pointing them to both the WGSIP and CMIP5 websites

to  feature an CMIP5-decadal link on the WGSIP (and perhaps other WGs’) front page(s) leading to some basic information and a link to the CMIP5 website

to note the existence of the CMIP5-decadal (i.e. CDFP) Panel as a contact point for questions in this area (at the moment Karl appears as the contact for all of CMIP5 on various web sites)

WGSIP may also discuss how we might contribute to  decadal studies in the areas of predictability, initialization, analysis and verification

Information you think the WG should be aware of or particular action they should consider taking?

WGCM (Meehl) suggests important activities in conjunction with WGSIP:

suggests WGSIP can provide guidance on how to evaluate the decadal hindcasts, experience in how to use hindcasts to

evaluate predictions in the SI community ENSO hindcast activities as a possible

template.   important input for Ch 11 of the AR5

notes potential Aspen Workshop next summer: synthesis paper on decadal hindcasts-

predictions for assessment by Ch 11, IPCC

CMIP5 (Taylor) responds data questions and other technical questions

mainly for CMIP5 will forward “science” questions he receives to

Panel (Latif: thanks for update … was panel officially

established?)

Possible follow up … Survey WGSIP subgroup on

“standard” analysis/verification applications (?) ….

offer input to Aspen workshop perhaps …

WGSIP and the IPCC

WGSIP recommendations for Chapter 11?

Title: Near-term climate prediction

long timescales in climate system forced and internally generated

predictability studies and results predictability as a feature of the physical system "skill" as our ability to correctly predict review of decadal scale predictability studies and implications

decadal forecast skill past studies but especially from CMIP5-decadal (CDFP) forecast combination, calibration, verification skill measures (deterministic and probabilistic) 

predictions 2015-25 etc.  (NOT projections dammit) from CMIP5-decadal deterministic and probabilistic in the context of past skill naturally includes regions - not a separate topic      

summary

WGSIP recommendations for Chapter 11?

The outline's last three topics:

Atmospheric composition and air quality Possible effects of geoengineering Quantification of the range of climate change projections

We will have few if any near term “forecasts" or "predictability studies" of:

atmospheric composition and air quality effects of geoengineering

The "quantification" of decadal predictions (NOT long term projections) is already, and properly, part of the "forecast skill"

don't need it as a separate topic don't want to make claims about Chapter 12 far-term climate projections

Does WGSIP have (should WGSIP have) recommendations for the organization of Chapter 11?

WGSIP decadal actions? survey community

as to status of CHFP and CDFP post results on website data, workshop implications for CHFP

CLIVAR site focus on “CDFP” links with and to WGSIP, WGCM, WGOMD sites (note CMIP5-decadal Panel as a contact point)

make comments/recommendations as to: methods, validation, evaluation, scores etc. IPCC Chapter 11 (topics, treatment..)

end of presentation

• should we do this?• other questions or or keep it simple?

Summary• Decadal prediction a natural extension of WGSIP’s SIP interests and of the

Climate-system Historical Forecast Project (CHFP)– IPCC class models– initialization of coupled system– retrospective forecasting of coupled system– focus on prediction and skill

• WGSIP/WGCM/CMIP5 partnership in “near term climate prediction” experiments

– complementary expertise– expanded community of models– joint oversight group

• Where are we in WGSIP’s view– workshops/projects are addressing some basic aspects – CMIP provides infrastructure support– need for forcing data – WGSIP/WGCM collaboration– reasonable progress in prospect– anticipate further workshops (e.g dealing with results; combination, calibration,

verification, implications for climate, application, value, …)

Other aspects; can WGSIP usefully address particular aspects of for instance:

retrospective forecast methodology initialization, ensemble generation forecast combination, calibration,

skill measures ….I.E. what is missing from meetings

and workshops?

Predictability and prediction predictability

a characteristic of a physical system itself a measure of the rate of separation of initially close states indicates the possibility of prediction

prognostic predictability studies typically use model to simulate “rate of separation” presumption that the model “similar enough” to real system

forecast skill characterized by error growth rate, decorrelation, or other

measure (rate of separation of actual and predicted states) indicates the current ability to predict

“potential predictability” analysis of variance; measure of signal to noise identifies regions where decadal variability is a useful

fraction of the total variance meant to indicate that prediction is potentially possible

Key issues - deBilt