Kelvin K. DroegemeierKelvin K. DroegemeierSchool of MeteorologySchool of MeteorologyUniversity of OklahomaUniversity of Oklahoma

AAAS Annual MeetingAAAS Annual Meeting15 February, 200915 February, 2009

Transforming Severe Transforming Severe Weather Prediction Through Weather Prediction Through

Dynamic AdaptationDynamic Adaptation

We Live in a Highly We Live in a Highly Vulnerable World…Vulnerable World…

……That Requires AdaptationThat Requires Adaptation

Observing Systems Do Not Observing Systems Do Not Sample the Atmosphere Sample the Atmosphere When/Where Needs are When/Where Needs are

GreatestGreatest

Forecast Models Run Forecast Models Run Largely on Fixed Largely on Fixed Schedules in Fixed Schedules in Fixed

DomainsDomains

The Nation’s The Nation’s Cyberinfrastructure is Cyberinfrastructure is

Virtually StaticVirtually Static

We Teach Using Current Weather We Teach Using Current Weather Data But Students Don’t Data But Students Don’t

InteractInteract With It With It

Why is Meteorology the Why is Meteorology the Exception?Exception?

A Fundamental Research A Fundamental Research QuestionQuestion

Can we better understand the atmosphere, Can we better understand the atmosphere, educate more effectively about it, and educate more effectively about it, and forecast more accurately if we forecast more accurately if we adaptadapt our our technologies and approaches to the weather technologies and approaches to the weather as it occursas it occurs??

People, even animals adapt/respond: Why People, even animals adapt/respond: Why don’t our resources???don’t our resources???

Sponsored by the National Science Foundation

The LEAD VisionThe LEAD VisionRevolutionize the ability of scientists, Revolutionize the ability of scientists,

students, and operational students, and operational practitioners to observe, analyze, practitioners to observe, analyze, predict, understand, and respond to predict, understand, and respond to intense local weather by interacting intense local weather by interacting with it with it dynamically and adaptivelydynamically and adaptively in in

real timereal time

Making it HappenMaking it Happen

Adaptive weather toolsAdaptive weather tools Adaptive sensorsAdaptive sensors Adaptive cyberinfrastructureAdaptive cyberinfrastructure

In a User-CenteredIn a User-CenteredFrameworkFrameworkWhere Where EverythingEverythingCanCanMutually InteractMutually Interact

                             

                                                               

Copyright © 2003 WGN-TV

Computer ModelsComputer Models are the Primary are the Primary Source of Information for All Source of Information for All Weather & Climate PredictionsWeather & Climate Predictions

The Prediction ProcessThe Prediction Process

Analyze ResultsAnalyze Results

Com

pare

and

Ver

ify

Com

pare

and

Ver

ify

Observe the AtmosphereObserve the Atmosphere

Identify and ApplyIdentify and ApplyPhysical LawsPhysical Laws

Create a MathematicalCreate a MathematicalModelModel

Create and Run aCreate and Run aComputer ModelComputer Model

The Prediction ProcessThe Prediction Process

Analyze ResultsAnalyze Results

Com

pare

and

Ver

ify

Com

pare

and

Ver

ify

Observe the AtmosphereObserve the Atmosphere

Identify and ApplyIdentify and ApplyPhysical LawsPhysical Laws

Create a MathematicalCreate a MathematicalModelModel

Create and Run aCreate and Run aComputer ModelComputer Model

Observe the Atmosphere Observe the Atmosphere

Upper-AirUpper-AirBalloonsBalloons

SatellitesSatellites

NEXRAD NEXRAD Doppler Doppler

RadarRadar

Commercial AircraftCommercial Aircraft

AutomatedAutomatedSurface Surface

NetworksNetworks

The Prediction ProcessThe Prediction Process

Analyze ResultsAnalyze Results

Com

pare

and

Ver

ify

Com

pare

and

Ver

ify Identify and ApplyIdentify and Apply

Physical LawsPhysical Laws

Create a MathematicalCreate a MathematicalModelModel

Create and Run aCreate and Run aComputer ModelComputer Model

Observe the AtmosphereObserve the Atmosphere

Identify & Apply Identify & Apply Physical LawsPhysical Laws

F=ma

The Prediction ProcessThe Prediction Process

Analyze ResultsAnalyze Results

Com

pare

and

Ver

ify

Com

pare

and

Ver

ify

Create a MathematicalCreate a MathematicalModelModel

Create and Run aCreate and Run aComputer ModelComputer Model

Observe the AtmosphereObserve the Atmosphere

Identify and ApplyIdentify and ApplyPhysical LawsPhysical Laws

Create a Mathematical ModelCreate a Mathematical Model

                                                                   

The Prediction ProcessThe Prediction Process

Analyze ResultsAnalyze Results

Com

pare

and

Ver

ify

Com

pare

and

Ver

ify

Create and Run aCreate and Run aComputer ModelComputer Model

Observe the AtmosphereObserve the Atmosphere

Identify and ApplyIdentify and ApplyPhysical LawsPhysical Laws

Create a MathematicalCreate a MathematicalModelModel

Create Computer Model Create Computer Model

Solve highly nonlinear partial differential Solve highly nonlinear partial differential equations equations

East/West WindEast/West Wind North/South WindNorth/South Wind Vertical WindVertical Wind TemperatureTemperature Water VaporWater Vapor Cloud WaterCloud Water Precipitating WaterPrecipitating Water Cloud IceCloud Ice GraupelGraupel HailHail Surface TemperatureSurface Temperature Surface MoistureSurface Moisture Soil TemperatureSoil Temperature Soil MoistureSoil Moisture Sub-Grid TurbulenceSub-Grid Turbulence

Run the Computer Model Run the Computer Model

Over the course of a single Over the course of a single forecast, the computerforecast, the computermodel solves billions of model solves billions of equationsequations

Requires the fastest Requires the fastest supercomputers in the supercomputers in the world -- capable ofworld -- capable ofperforming quadrillionsperforming quadrillionsof calculationsof calculationseach secondeach second

Run the Computer Model Run the Computer Model

The Prediction ProcessThe Prediction Process

Analyze ResultsAnalyze Results

Com

pare

and

Ver

ify

Com

pare

and

Ver

ify

Observe the AtmosphereObserve the Atmosphere

Identify and ApplyIdentify and ApplyPhysical LawsPhysical Laws

Create a MathematicalCreate a MathematicalModelModel

Create and Run aCreate and Run aComputer ModelComputer Model

Computer Weather Prediction Began Computer Weather Prediction Began with a Vision in the early 1920s -- with a Vision in the early 1920s --

L.F. Richardson’s “Forecast L.F. Richardson’s “Forecast Factory”Factory”

Richardson’s Forecast Grid – Richardson’s Forecast Grid – Predictions Done by HandPredictions Done by Hand

25 point mesh!

One Level

Grid Spacing = 250 km

The Vision Becomes Reality… The Vision Becomes Reality… ENIACENIAC

ENIAC Versus TodayENIAC Versus Today

Weighed 30 tonsWeighed 30 tons Had 18,000 vacuum tubes, 1,500 Had 18,000 vacuum tubes, 1,500 relays thousands of resistors, relays thousands of resistors, capacitors, inductorscapacitors, inductors

Peak speed of 5000 adds/second and Peak speed of 5000 adds/second and 300 multiplies/sec300 multiplies/sec

A 1.2 GHz Pentium IV processor is A 1.2 GHz Pentium IV processor is 500,000 times faster than the 500,000 times faster than the ENIACENIAC

A desktop PC with 1 Gbyte of RAM A desktop PC with 1 Gbyte of RAM can store 5 million times as much can store 5 million times as much data as the ENIAC data as the ENIAC

Numerically Numerically integrated integrated oneone equation at equation at oneone altitudealtitude

736 km grid spacing736 km grid spacing 24 hour forecast 24 hour forecast took 24 hours to took 24 hours to compute!compute!

Forecast blew up Forecast blew up due to lack of due to lack of smoothing of data – smoothing of data – but rerun today, it but rerun today, it was ok!was ok!

450 Miles

1950: The First 1950: The First ComputerComputer Weather Forecast ModelWeather Forecast Model

Today’s ModelsToday’s Models

Typical Forecast from Today’s Typical Forecast from Today’s Operational ModelsOperational Models

Why the Lack of Detail in Why the Lack of Detail in the Model? the Model?

This ThunderstormFalls Through the Cracks

Why the Lack of Detail in Why the Lack of Detail in the Model? the Model?

What Causes the Major What Causes the Major Problems?Problems?

A Foundational QuestionA Foundational Question

. . . explicitly predict this. . . explicitly predict thistype of weather?type of weather?

Can computer forecastCan computer forecasttechnology. . .technology. . .

The Answer is Yes…Sort The Answer is Yes…Sort Of!Of!

Tornado

NWS 12-hr Computer Forecast Valid at 6 pm CDTNWS 12-hr Computer Forecast Valid at 6 pm CDTNo No Explicit EvidenceExplicit Evidence of Precipitation in of Precipitation in

North TexasNorth Texas

Reality Was Quite Different!Reality Was Quite Different!

6 pm 7 pm 8 pmR

adar

Xue et al. (2003)

Fort Worth

Fcs

t W

ith

Rad

ar D

ata

2 hr 3 hr 4 hr

Fort Worth

As a As a Forecaster Forecaster

Worried About Worried About This Reality… This Reality…

7 pm

As a As a Forecaster Forecaster

Worried About Worried About This Reality… This Reality…

How Much How Much Trust Would Trust Would You Place in You Place in This Model This Model Forecast? Forecast?

3 hr

7 pm

Initial State Uncertainty

Truth

Single Forecast

Traditional Forecasting

Methodology

t critical

Deterministic Forecast

Probabilistic Forecast

Ensemble Forecasting

Initial State Uncertainty

Mean

Truth

Actual RadarActual Radar

Ensemble Member #1Ensemble Member #1 Ensemble Member #2Ensemble Member #2

Ensemble Member #3Ensemble Member #3 Ensemble Member #4Ensemble Member #4Control ForecastControl Forecast

Actual RadarActual Radar

Probability of Intense Probability of Intense PrecipitationPrecipitation

Model Forecast Radar Observations

Now I need a LOT More Now I need a LOT More Computing CapacityComputing Capacity

Can We Do Better if we Adapt Can We Do Better if we Adapt Rather than React??Rather than React??

Charles Darwin

Sample Problem Scenario in Sample Problem Scenario in AdaptationAdaptation

StreamingObservations

Storms Forming

Forecast Model

On-DemandGrid Computing

Data Mining

How Does LEAD Do It? How Does LEAD Do It? The Notion of a Web ServiceThe Notion of a Web Service

Web Service: A Web Service: A program that carries program that carries out a specific set out a specific set of operations based of operations based upon upon requests from requests from clientsclients

The LEAD The LEAD architecture is a architecture is a “Service Oriented “Service Oriented Architecture”Architecture” (SOA), (SOA), which means that all which means that all of the key functions of the key functions are represented as a are represented as a set of servicesset of services. .

Service-Oriented ArchitectureService-Oriented Architecture

Service A(Analysis)

Service B(Model)

Service C(Radar Stream)

Service D(Work Space)

Service E(VO Catalog)

Service F(Viz Engine)

Service G(Monitoring)

Service H(Scheduling)

Service I(Decoder)

Service J(Repository)

Service K(Mining)

Service L(Decoder)

Many others…

Service B(Model)

Service A(Analysis)

Service C(Radar Stream)

Service D(Work Space)

Service K(Mining)

Service L(Decoder)

Service J(Repository)

Can Solve Broad Classes of ProblemsCan Solve Broad Classes of Problemsby Linking Services Together in by Linking Services Together in WorkflowsWorkflows

Experiment Builder

Research to Operational Research to Operational Practice: NOAA Hazardous Practice: NOAA Hazardous

Weather Test BedWeather Test Bed LEAD produced on-demand LEAD produced on-demand forecasts for experimental forecasts for experimental evaluation by operational evaluation by operational forecasters at the National forecasters at the National Storm Prediction CenterStorm Prediction Center– Fully automatedFully automated– Forecaster-initiatedForecaster-initiated

Mid-April – early June (severe Mid-April – early June (severe weather seasons) 2006, 2007, weather seasons) 2006, 2007, 20082008

The Value of Adaptation: The Value of Adaptation: Forecaster-Initiated Forecaster-Initiated

Predictions on 7 June 2007Predictions on 7 June 2007

Brewster et al. (2008)

Radar Observations Standard 20-hr Forecast5 hr LEAD Dynamic Forecast

Centers of On-Demand Forecast Grids Centers of On-Demand Forecast Grids Launched at NCSA During 2007 Spring Launched at NCSA During 2007 Spring

ExperimentExperiment

Launched automatically in response to hazardous weather messages (tornado watches, mesoscale discussions)

Launched based on forecaster guidance

Graphic Courtesy Jay Alameda and Al Rossi, NCSA

The Million Dollar The Million Dollar Question: Will Question: Will Computer Models Computer Models Ever Be Able to Ever Be Able to

PredictPredict Tornadoes? Tornadoes?

Schematic Diagram of a Supercell Storm (C. Doswell)

Predicting Tornadoes: Predicting Tornadoes: The Warn on Forecast The Warn on Forecast

ConceptConcept

We’re Capturing the We’re Capturing the Attention of Key Attention of Key

Leaders!!Leaders!!