using high-performance computing to predict extreme weather
TRANSCRIPT
![Page 1: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/1.jpg)
October 29, 2014
Using High-Performance
Computing to Predict Extreme
WeatherPeter Bauer
![Page 2: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/2.jpg)
October 29, 2014October 29, 2014
Independent intergovernmental organisation
established in 1975
with19 Member States15 Co-operating States
European Centre for Medium-Range Weather Forecasts
![Page 3: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/3.jpg)
October 29, 2014
The success story of Numerical Weather Prediction: Wind storms28-29 October 2013 ‘Christian’ wind gusts
5-day forecast of probabilities for wind gusts exceeding > 10m/s
![Page 4: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/4.jpg)
October 29, 2014
The success story of Numerical Weather Prediction: Heat wavesMaximum observed temperature
Maximum forecast temperature day-3 – day-4
Temperature anomaly ensemble forecast week-2
… and week-3½
Week 2-3½ forecast of probabilities for temperature anomalies exceeding > 5-10K
![Page 5: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/5.jpg)
October 29, 2014
29-09-15 30-09-15 01-10-15
Individual trajectories for JOAQUIN during the next 240 hourstracks: thick solid=HRES; thick dot=CTRL; thin solid=EPS members [coloured]
The success story of Numerical Weather Prediction: Hurricanes
![Page 6: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/6.jpg)
October 29, 2014
Day 0
Day 2Southern Gaza Strip, on April18, 2012
Day 4
The success story of Numerical Weather Prediction: Dust storms
![Page 7: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/7.jpg)
October 29, 2014
Previously on …?
John von Neumann (1903-1957), Mathematician:• Function theory, abstract algebra,
quantum physics• Leader of Electronic Computer Project
(1946-52)
Jule Charney (1917-1981), Meteorologist:• Set of equations for numerical prediction of planetary waves• Founder of theory of baroclinic instability
Electronic Numerical Integrator and Computer (ENIAC)• 140 kW, 30 tons, 18,000 thermo-ionic valves• 1-layer model, resolution 400-700 km, North American domain• Single 24-hour forecast needed 24 hours compute time
ENIAC 1950
The same prediction needed 1 second on a Nokia 6300 mobile phone (2006)!
Lewis Fry Richardson (1881-1953), Physicist, Meteorologist, Psychologist, Pacifist:• Basics of numerical weather prediction• First explicit calculation of weather on 20 May 1910
![Page 8: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/8.jpg)
October 29, 2014
Computer power
Model complexity
Model resolution
Tomorrow
Today
Multiple dimensions
EnsemblesLong climate runs
![Page 9: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/9.jpg)
October 29, 2014
What is the challenge?
Observations Models
Volume 20 million = 2 x 107 5 million grid points100 levels10 prognostic variables = 5 x 109
Type 98% from 60 different satellite instruments
physical parameters of atmosphere, waves, ocean
Observations Models
Volume 200 million = 2 x 108 500 million grid points200 levels100 prognostic variables = 1 x 1013
Type 98% from 80 different satellite instruments
physical and chemical parameters of atmosphere, waves, ocean, ice, vegetation
Today:
Tomorrow:
Factor 10 Factor 2000per day per time step
(10-day forecast = 1440 time steps)
![Page 10: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/10.jpg)
October 29, 2014
scalability range
scalability range
Ensemble
Single
Simple compute projection (only resolution)
2015 2025
M€ electricity/year
Power limit
[Bauer et al. 2015]
![Page 11: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/11.jpg)
October 29, 2014
2015 Time critical• 21 TB/day written• 22 million fields• 85 million products• 11 TB/day sent to customers
Non-time critical• 100 TB/day archived• 400 research experiments• 400,000 jobs / day
20202025?
Time critical• 128 TB/day written• 90 million fields• 450 million products• 60 TB/day sent to customers
Non-time critical• 1 PB/day archived• 1,000 research experiments• 1,000,000 jobs / day
Data projection
Factor 5-10 every 5 years!
time
![Page 12: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/12.jpg)
October 29, 2014
[Schulthess 2015]
Traditional science workflow
![Page 13: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/13.jpg)
October 29, 2014
[Schulthess 2015]
Future science workflow
science specific code generic code
Energy efficient SCalable Algorithms for weather Prediction at Exascale www.hpc-escape.eu
![Page 14: Using High-Performance Computing to Predict Extreme Weather](https://reader035.vdocuments.site/reader035/viewer/2022070601/588a50b01a28abd3088b6c81/html5/thumbnails/14.jpg)
October 29, 2014October 29, 2014
‘The quiet revolution of numerical weather prediction’ by P. Bauer, A.J. Thorpe, G. Brunet
in Nature, 3 September 2015:
The future of Numerical Weather Prediction
[IPCC]