Kelvin K. Droegemeier School of Meteorology University of Oklahoma AAAS Annual Meeting 15 February, 2009 Transforming Severe Weather Prediction Through Dynamic Adaptation
We Live in a Highly Vulnerable World…
…That Requires Adaptation
Observing Systems Do Not Sample the Atmosphere When/Where Needs are Greatest Forecast Models Run Largely on Fixed Schedules in Fixed Domains The Nation’s Cyberinfrastructure is Virtually Static We Teach Using Current Weather Data But Students Don’t Interact With It Why is Meteorology the Exception?
A Fundamental Research Question A Fundamental Research Question n Can we better understand the atmosphere, educate more effectively about it, and forecast more accurately if we adapt our technologies and approaches to the weather as it occurs? n People, even animals adapt/respond: Why don’t our resources???
Sponsored by the National Science Foundation
The LEAD Vision Revolutionize the ability of scientists, students, and operational practitioners to observe, analyze, predict, understand, and respond to intense local weather by interacting with it dynamically and adaptively in real time
Making it Happen n Adaptive weather tools n Adaptive sensors n Adaptive cyberinfrastructure n In a User-Centered n Framework n Where Everything n Can n Mutually Interact
Copyright © 2003 WGN-TV Computer Models are the Primary Source of Information for All Weather & Climate Predictions
The Prediction Process Analyze Results Compare and Verify Observe the Atmosphere Identify and Apply Physical Laws Create a Mathematical Model Create and Run a Computer Model
The Prediction Process Analyze Results Compare and Verify Observe the Atmosphere Identify and Apply Physical Laws Create a Mathematical Model Create and Run a Computer Model
Observe the Atmosphere Upper-AirBalloons Satellites NEXRADDopplerRadar Commercial Aircraft AutomatedSurfaceNetworks
The Prediction Process Analyze Results Compare and Verify Identify and Apply Physical Laws Create a Mathematical Model Create and Run a Computer Model Observe the Atmosphere
Identify & Apply Physical Laws F=ma
The Prediction Process Analyze Results Compare and Verify Create a Mathematical Model Create and Run a Computer Model Observe the Atmosphere Identify and Apply Physical Laws
Create a Mathematical Model
The Prediction Process Analyze Results Compare and Verify Create and Run a Computer Model Observe the Atmosphere Identify and Apply Physical Laws Create a Mathematical Model
Create Computer Model
n Solve highly nonlinear partial differential equations n East/West Wind n North/South Wind n Vertical Wind n Temperature n Water Vapor n Cloud Water n Precipitating Water n Cloud Ice n Graupel n Hail n Surface Temperature n Surface Moisture n Soil Temperature n Soil Moisture n Sub-Grid Turbulence Run the Computer Model
n Over the course of a single forecast, the computer model solves billions of equations n Requires the fastest supercomputers in the world -- capable of performing quadrillions of calculations each second Run the Computer Model
The Prediction Process Analyze Results Compare and Verify Observe the Atmosphere Identify and Apply Physical Laws Create a Mathematical Model Create and Run a Computer Model
Computer Weather Prediction Began with a Vision in the early 1920s -- L.F. Richardson’s “Forecast Factory”
Richardson’s Forecast Grid – Predictions Done by Hand 25 point mesh! One Level Grid Spacing = 250 km
The Vision Becomes Reality… ENIAC
ENIAC Versus Today n Weighed 30 tons n Had 18,000 vacuum tubes, 1,500 relays thousands of resistors, capacitors, inductors n Peak speed of 5000 adds/second and 300 multiplies/sec n A 1.2 GHz Pentium IV processor is 500,000 times faster than the ENIAC n A desktop PC with 1 Gbyte of RAM can store 5 million times as much data as the ENIAC
n Numerically integrated one equation at one altitude n 736 km grid spacing n 24 hour forecast took 24 hours to compute! n Forecast blew up due to lack of smoothing of data – but rerun today, it was ok! 450 Miles 1950: The First Computer Weather Forecast Model
Today’s Models
Typical Forecast from Today’s Operational Models
Why the Lack of Detail in the Model? This Thunderstorm Falls Through the Cracks
Why the Lack of Detail in the Model?
What Causes the Major Problems?
A Foundational Question... explicitly predict this type of weather? Can computer forecast technology...
The Answer is Yes…Sort Of!
Tornado
NWS 12-hr Computer Forecast Valid at 6 pm CDT No Explicit Evidence of Precipitation in North Texas
Reality Was Quite Different!
6 pm 7 pm8 pm Radar Xue et al. (2003) Fort Worth Fcst With Radar Data 2 hr 3 hr 4 hr Fort Worth
As a Forecaster Worried About This Reality… 7 pm
As a Forecaster Worried About This Reality… How Much Trust Would You Place in This Model Forecast? 3 hr 7 pm
Actual Radar
Ensemble Member #1 Ensemble Member #2 Ensemble Member #3 Ensemble Member #4 Control Forecast Actual Radar
Probability of Intense Precipitation Model Forecast Radar Observations
Now I need a LOT More Computing Capacity
Can We Do Better if we Adapt Rather than React?? Charles Darwin
Sample Problem Scenario in Adaptation Streaming Observations Storms Forming Forecast Model On-Demand Grid Computing Data Mining
How Does LEAD Do It? The Notion of a Web Service n Web Service: A program that carries out a specific set of operations based upon requests from clients n The LEAD architecture is a “Service Oriented Architecture” (SOA), which means that all of the key functions are represented as a set of services.
Service-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 Problems by Linking Services Together in Workflows
Experiment Builder
Research to Operational Practice: NOAA Hazardous Weather Test Bed n LEAD produced on-demand forecasts for experimental evaluation by operational forecasters at the National Storm Prediction Center –Fully automated –Forecaster-initiated n Mid-April – early June (severe weather seasons) 2006, 2007, 2008
The Value of Adaptation: Forecaster- Initiated Predictions on 7 June 2007 Brewster et al. (2008) Radar Observations Standard 20-hr Forecast 5 hr LEAD Dynamic Forecast
Centers of On-Demand Forecast Grids Launched at NCSA During 2007 Spring Experiment 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 Question: Will Computer Models Ever Be Able to Predict Tornadoes?
n Schematic Diagram of a Supercell Storm (C. Doswell)
Predicting Tornadoes: The Warn on Forecast Concept
We’re Capturing the Attention of Key Leaders!!