Forecast development at the IRI

• Michael K. Tippett

Our approachVisionValues

Past, present and future strategies

Outline

VisionProvide global climate forecasts for societal benefit

ValuesHigh-quality climate forecast information/ingredients

In-housePartners

Transform forecast information into useful productsMatch user systems with scientific capabilities

Products informed by research (R2O/O2R)PhysicalSocial

Provide benefit via product content as well as process“Best practices”

Approach to forecast development

One that is used.Fits into a user’s decision system

Hard to convince users to change their systemsEasier to get them to add inputs

Available.Data library

Trusted.Verification

What is a good forecast product?

Forecast verification

Forecast inputs/ingredientsForecast modelsObservational data

ProductsCategorical probabilitiesPDFs

MethodologiesCombination/calibration of forecast inputsProduct delivery

Strategic elements

Carbon

Ocean

AtmosphereLand

Chemistry

Ice

Differing classes of forecast models

http://www.cmmap.org

Two classes of forecast models:• Ocean-atmosphere coupled

models: Initial state of climate system is prescribed

• Atmosphere-only models: Future SST is prescribed

Coupled processes

The very beginning

Forecast ingredients:• Prescribed SST AGCMs (not coupled)Products• Issued seasonally• 3-month averages• Near-surface temperature and precipitation• Tercile probabilities• No digital dataMethodology• Basis for RCOF (subjective)• Manual map production

ATB (After Tony Barnston)

Forecast ingredients:• Prescribed SST AGCMs (not coupled)Products• Issued monthly• 3-month averages• Near-surface temperature and precipitation• Tercile probabilities• Digital data availableMethodology• Objective estimation of probabilities• Automated map production (CRED)

Present

Forecast ingredients:• Prescribed AGCMs• CFSv2 (coupled)Products• Issued monthly• 3-month averages• Near-surface temperature and precipitation• Maps of tercile probabilities• Full PDFs• Digital data via Data LibraryMethodology• Objective estimation of probabilities• Automated map production• More realistic estimates of uncertainty

Flexible forecast format maproom

Forecast inputsMore coupled modelsNMME

ProductsAdditional quantities, time-scalesLeverage emerging research

MethodologiesMore agile, able to adapt to changing inputsLeverage emerging research

Future

Forecast and monitoring of regional extremes

Observations

Forecasts

Verification timeFo

reca

st le

ad (d

ays)

Monitor and forecastregional indices e.g.:• Rainfall• Severe weather• Fire

Motivated by IRFC collaboration

0-lead

45-lead