Скачать презентацию An Inside Look at CPC s Medium and Long Скачать презентацию An Inside Look at CPC s Medium and Long

b6d6e4bd3b39a458f221b650f5ce7329.ppt

  • Количество слайдов: 48

An Inside Look at CPC’s Medium and Long. Range Forecasts Ed O’Lenic NOAA-NWS-Climate Prediction An Inside Look at CPC’s Medium and Long. Range Forecasts Ed O’Lenic NOAA-NWS-Climate Prediction Center Camp Springs, Maryland ed. olenic@noaa. gov 301 -763 -8000, ext 7528

WEATHER vs. CLIMATE • Wildly oscillating curve = daily “weather” • Smooth curve = WEATHER vs. CLIMATE • Wildly oscillating curve = daily “weather” • Smooth curve = 30 year mean (climatology)

Forecast Process Schematic Dynamical model forecasts/multimodel ensembles Recent observations Historical observations. . Verifications/Statistical tools Forecast Process Schematic Dynamical model forecasts/multimodel ensembles Recent observations Historical observations. . Verifications/Statistical tools Downscaling, Analogs, Composites WEB PAGES/AUTOMATED DATABASES Peer-reviews of the forecast tools and of the penultimate forecast via web/telephone conference with partners and through local discussions (map discussions, sanity check, conference calls, etc…) Forecaster-created or automated products Dissemination to public

Generic Seasonal Climate Forecast Map 5 0 EC B 10 5 0 A EC Generic Seasonal Climate Forecast Map 5 0 EC B 10 5 0 A EC EC=Equal Chances for the tercile categories, 33 -1/3 each. Contours are labeled with the deviation from EC for the indicated category.

Part 1. Long-Lead Seasonal Forecasts Part 1. Long-Lead Seasonal Forecasts

Forecast Maps and Bulletins • Each month, on the Thursday between the 15 th Forecast Maps and Bulletins • Each month, on the Thursday between the 15 th 21 st, CPC issues a set of 13 seasonal outlooks. • There are two maps for each of the 13 leads, one for temperature and one for precipitation for a total of 26 maps. • Each outlook covers a 3 -month “season”, and each forecast overlaps the next and prior season by 2 months. • Bulletins include: the prognostic discussion for the seasonal outlook over North America, and, for Hawaii. • The monthly outlook is issued at the same time as the seasonal outlook. It consists of a temperature and precipitation outlook for a single lead, 0. 5 months, and the monthly bulletin. • All maps are sent to AWIPS, Family of Services and internet.

Statistical Prediction Tools • Multiple Linear Regression: - Predicts a single variable from historical Statistical Prediction Tools • Multiple Linear Regression: - Predicts a single variable from historical and recent observations of two or more predictors. • Canonical Correlation Analysis (CCA): – Uses recent and historical observations of Northern Hemisphere circulation (Z), global sea surface temperature (SST), US surface T (Tus) to create a set of 5 or 6 EOFs of predictors and predictands. – Looks at cross-correlations between time series of predictors and predictands. – Predicts temporal and spatial patterns from patterns.

Statistical Prediction Tools • Constructed Analogs (CA) – Uses recent observations (base) of a Statistical Prediction Tools • Constructed Analogs (CA) – Uses recent observations (base) of a single variable and historical observations, to construct a weighted mean of all prior years which best explains the base data. Assumes the evolution to subsequent seasons is also best explained by the weights used to construct the analog to the base. • Optimal Climate Normals (OCN) – Uses the difference between the most recent 10 (15) years of temperature (precipitation) observations and the 30 -year climatology (i. e. , the trend) for a given season as the prediction for future occurrences of that season.

Detailed operations concept for NCEP’s ocean -atmosphere model An ensemble of 16 ocean SST Detailed operations concept for NCEP’s ocean -atmosphere model An ensemble of 16 ocean SST forecasts are created using a coupled GCM. The average of these is used as the lower boundary for… An AGCM, along with 20 different sets of initial conditions, to create a set of 20 ensemble atmosphere forecasts out to 9 months. A 20 -year AMIP run of the AGCM is made each month for use as the climatology to create anomalies/remove model bias. In collaboration with partners (CDC, IRI), forecasters use the NCEP model tools, together with other model tools to subjectively create outlook maps of the probability of monthly and seasonal mean temperature and total precipitation category.

NCEP Two-Tier Climate Modeling System SST TOPEX XBT TAO INTEGRATED OCEAN MODELDATA ASSIMILATION SYSTEM NCEP Two-Tier Climate Modeling System SST TOPEX XBT TAO INTEGRATED OCEAN MODELDATA ASSIMILATION SYSTEM OCEAN INITIAL CONDITIONS AGCM FORECAST S COUPLED OCEANATMOSPHERE GCM CDC IRI, CDC SURFACE T, P ANOMALIES SSTA STRESS EVAPPRECIP FLUX HEAT FLUXES STATISTICAL TOOLS: CCA, CA SSMI/ERS-2 OFFICIAL SST FCST FORECASTERS OFFICIAL PROBABILISTIC T, P OUTLOOKS

Forecast tools web page Forecast tools web page

Global SST Global SST

TAO Ocean T Obs TAO Ocean T Obs

NCEP AGCM Forecasts for DJF 2000 -01, 2001 -02, and 2002 -03 Global and NCEP AGCM Forecasts for DJF 2000 -01, 2001 -02, and 2002 -03 Global and NOAM T Fcst SST Forcing 2000 -01 2001 -02 2002 -03

MAM 2003 NCEP AGCM T Forecast MAM 2003 NCEP AGCM T Forecast

CCA 0. 5 Mo lead MAM T Outlook CCA 0. 5 Mo lead MAM T Outlook

OCN 0. 5 Mo lead MAM T Outlook OCN 0. 5 Mo lead MAM T Outlook

ENSO Composites given CPC consolidated SST forecast ENSO Composites given CPC consolidated SST forecast

OFFICIAL MAM 2003 T Outlook OFFICIAL MAM 2003 T Outlook

U. S. Temperature Skill SS 1= ((c-e)/(t-e))*100 Mean ss 1 Mean ss 2 0 U. S. Temperature Skill SS 1= ((c-e)/(t-e))*100 Mean ss 1 Mean ss 2 0 SS 2= ((c+(1/3)*cl - e)/(t-e))*100

Precipitation Skill Precipitation Skill

The Forecast • 1 -3 years: Week 2 forecasts for the Pacific Region & The Forecast • 1 -3 years: Week 2 forecasts for the Pacific Region & Caribbean based on MJO, ENSO, Monsoon, improved dynamical & statistical models. • 3 -5 years: Week 3, 4 forecast based on NAM/AO relationships, MJO, ENSO, improved statistical & dynamical models. • 5 -10 years: Improved seasonal, monthly, week 2, 3 -4 forecasts based on improved dynamical & statistical model prediction of NAM/AO, diurnal cycle of convection, MJO, ENSO, Monsoon, decadal oscillations, ocean-atmosphere coupling, MM ensembles, more & better observations, …

Part 2. Medium-Range Forecasts Part 2. Medium-Range Forecasts

6 -10 day/week 2 process schematic Multi-model ensemble R 9: 00 AM Weighted average 6 -10 day/week 2 process schematic Multi-model ensemble R 9: 00 AM Weighted average of model 500 h. Pa height Downscale: get surface weather from 500 mb height via analogs, regression, neural network. R Forecaster formulates maps of predicted T, P, PMD bulletin Disseminate via web, AWIPS, FOS 3 -4 PM R = Forecaster reconciliation of tools required

Recent Changes to Procedures • • • From 3 times/week to daily in October Recent Changes to Procedures • • • From 3 times/week to daily in October 2000 Automated weekend forecasts from October 2000 Percent probability format from October 2000 Alaska and week 2 added October 2000 Automated weekend forecasts improved October 2001—neural net tool omitted and consistency with weekday forecasts added • Bias-corrected precipitation forecast tool and other improvements added in the fall of 2001

Forecast Maps and Bulletins Each day, between 3 and 4 PM Eastern Time, CPC Forecast Maps and Bulletins Each day, between 3 and 4 PM Eastern Time, CPC issues a set of 6 -10 day and week 2 outlooks. These are formulated by a forecaster (Monday through Friday) and are automated on weekends. There are two 500 mb height maps, two surface maps and a single bulletin. Sample 6 -10 day outlook 500 mb height and anomaly forecast map from CPC web page.

GFS Ensemble upper-airheight forecasts, analog GFS Ensemble upper-airheight forecasts, analog

ECMWF upper-air-height forecasts, analog ECMWF upper-air-height forecasts, analog

Official 6 -10 day 500 h. Pa forecast Official 6 -10 day 500 h. Pa forecast

Teleconnections (TC) Definition: Composite of those maps, for a calendar month, with largest + Teleconnections (TC) Definition: Composite of those maps, for a calendar month, with largest + (top 10%) or – (bottom 10%) 500 h. Pa height at a specified space point from 19501999 (~150 maps). Forecaster computes TC on major anomaly centers (base points) of 500 h. Pa forecast maps. Strong TC ~ large correlation values at the distant centers ~ frequent/persistent pattern Weak TC ~ the pattern is probably transient and not as likely to be well predicted by the model as would a persistent pattern.

Teleconnection on 500 h. Pa center at 50 N/140 W (+) and 55 N/90 Teleconnection on 500 h. Pa center at 50 N/140 W (+) and 55 N/90 W (-)

Composite of observed T, P anomalies associated with teleconnecion on + 500 h. Pa Composite of observed T, P anomalies associated with teleconnecion on + 500 h. Pa anomalies at 56 N 10 W

ENSEMBLE T, P prediction analog maps ECMWF ENSEMBLE T, P prediction analog maps ECMWF

8 -14 Day MRF precipitation bias correction 6 -10 Day 8 -14 Day MRF precipitation bias correction 6 -10 Day

Official 6 -10 day T forecast Official 6 -10 day T forecast

Official 6 -10 day P forecast Official 6 -10 day P forecast

6 -10 day Monthly Average Skill Scores s=((c-e)/(t-e))*100 c = # hits e = 6 -10 day Monthly Average Skill Scores s=((c-e)/(t-e))*100 c = # hits e = # chance hits t = # forecasts

Skill of Official 6 -10 day T, P, 500 h. Pa Ensemble Mean Z Skill of Official 6 -10 day T, P, 500 h. Pa Ensemble Mean Z

Skill of 6 -10 day T tools, 500 h. Pa Ensemble Mean Z Skill of 6 -10 day T tools, 500 h. Pa Ensemble Mean Z

The Forecast • 1 -3 years: Week 2 forecasts for the Pacific Region & The Forecast • 1 -3 years: Week 2 forecasts for the Pacific Region & Caribbean based on MJO, ENSO, Monsoon, improved dynamical & statistical models. • 3 -5 years: Week 3, 4 forecast based on NAM/AO relationships, MJO, ENSO, improved statistical & dynamical models. • 5 -10 years: Improved seasonal, monthly, week 2, 3 -4 forecasts based on improved dynamical & statistical model prediction of NAM/AO, diurnal cycle of convection, MJO, ENSO, Monsoon, decadal oscillations, ocean-atmosphere coupling, MM ensembles, more & better observations, …

The End The End