The IPS Extreme Space Weather Service

Size: px

Start display at page:

Download "The IPS Extreme Space Weather Service"

Norah Leonard
5 years ago
Views:

1 The IPS Extreme Space Weather Service R. Marshall, M. Terkildsen, G. Steward, D. Neudegg, Z. Bouya Ionospheric Prediction Service (IPS) Australian Bureau of Meteorology w w w. i p s. g o v. a u

2 ESW Motivation Growing awareness of impacts of space weather has lead to increased interest from critical infrastructure groups within Australia Australian Government TISN, includes Energy, Communications, and Transport Sector groups Energy sector includes power network asset owners and operators, gas pipeline owners and operators, mining companies Communications sector includes, Telstra, Optus, major TV broadcasters Transport sector includes major airlines (Qantas, Virgin), ASA, shipping and rail

3 Satellite Systems CRESS Electon and Proton Fluxes (Brautigan, 2002) Chilean Ash Cloud, June 2011

Getley et al., Space Weather, 2004SW000110 Aviation International guidelines: 100 msv/yr over 5 yrs < 50 msv max any year < 1 msv pregnant women Oct 29, 2003 Lax - JFK 5 4.

4 Getley et al., Space Weather, 2004SW Aviation International guidelines: 100 msv/yr over 5 yrs < 50 msv max any year < 1 msv pregnant women Oct 29, 2003 Lax - JFK Kp = 8, 9 LAX-JFK QF107 29th Oct 2003 'A' Total Equivalent Dose = 12.0 Micro Sieverts Australian airline aircraft altered course during significant SPE event earlier this year GLE Dose µsv Micro Sieverts/hr 3 per Hr Total Dose 12 µsv :51 18:03 18:16 18:28 Alt 37,000 Time UTC UT 18:41 18:54 19:07 19:21 19:32 19:45 19:58 20:11 Alt to 39,000 20: :49 21:02 21:15 21:28 21:41 21:53 22:06 M. Duldig, adapted from Getley et al., Space Weather, 2004SW000110

5 Low-End GPS Systems (from

6 High-End GPS Systems (Terkildsen, 2010, IPS-CR P, April 2010) [from Pullen et al, 5th International GBAS Working Group Meeting (GWG/5), Nov 2006] [from Ene et al (2005), Proceedings of ION 2005 National Technical Meeting]

7 Power Networks Threat Level Model from Global network observations Marshall, Smith, Francis, Waters and Sciffer, Space Weather, 2011

Australian/NZ Power Networks 300 250 GIC y -index

, Space Weather, 2011SW000685 200 180 160 GIC y Index

1985 1990 1995 Time 2000 2005 2010 HBT CNB GNA ASP LRM

8 Australian/NZ Power Networks GIC y -index Occurrences > 50 Marshall et al., Space Weather, 2011SW GIC y Index Sunspot Number Time HBT CNB GNA ASP LRM CTA KDU Sunspot Marshall et al., Space Weather, 2012SW Copyright, Transpower, New Zealand

9 Extreme Space Weather (ESW) Model Most space weather impacts in Australian region associated with extreme events Generalised Linear Model (GLM) techniques Event-based analysis Requirements: LATENCY Long range warning (> 12 hours) Based on solar data only Short range alert (~ 1 hour) Based on solar data + ACE ACCURACY Long range: Optimise to minimise missed events Short range: Optimise for forecast accuracy SIMPLICITY Design for active use in space weather forecast environment

10 ESW Model Events Event- based analysis M1 Associated X-ray Solar IMF Major flare CME wind Bzstorm south specifies shock event solar disturbance event Flares / CMEs solar wind (shocks) IMF (Bz events) Dst (storm events) day of year

11 ESW Model Parameters Model covariates (the input data ) X-RAY FLARE Solar flare magnitude Solar flare duration LOCATION OF SOLAR ACTIVE REGION Latitude of solar active region Longitude of solar active region SOLAR CYCLE Solar Cycle Phase SOLAR WIND / IMF IMF Bz Solar wind shock CME CHARACTERISTICS Presence of Halo CME (CME width) CME speed

12 Generalised Linear Model Response variable (what is being modelled) dstn = 0 Dst + 49 ESW Model GLM Dst 50 else Model (a GLM) ln( µ ) = α + α x + α x + L+ α i i 0 0, pi = 1, pi 1 ( y µ ) ( y µ ) i1 Training data (for fitting model coefficients) y < p p 2 thresh thresh i2 15 years data ( ) Prediction (a binary output) m x im xi 1 Kx im p i ( y µ ) Input variables a 0 Ka m Coefficients (to fit) pthresh Response distribution Threshold on event probability used to produce binary prediction (ESW event/no-event). Optimised for required model performance.

13 ESW Model Validation Model validation: Solar data only Optimising for no missed events (false negatives = 0) Modelled event probability false positives (25) true positives (6) false negatives (0) Observed dstn Leave-One-Out Cross Validation (LOOCV) true positives 6 false positives 25 false negatives 0 total events 644 Probability threshold (to determine binary outcome) ESW event threshold (dst < -250)

and probabilistic forecasts for ESW Accepts a range of

14 ESW Model Implementation Operational GUI in ASFC Uses Dst-based ESW models as a back-end, proving both binary and probabilistic forecasts for ESW Accepts a range of covariates for added flexibility Simple text warning message

15 ACE solar wind parameters CME symmetry parameter / CME mass Active region magnetic characteristics (proxy for IMF Bz events?) CME travel time (flare-shock interval) to replace CME plane-of-sky speed (a poor proxy for true CME speed) Type II / Type IV radio bursts Direct modelling of ESW parameters as response GIC index Future Developments Ionospheric gradient index

16 w w w. i p s. g o v. a u Thank you.

Montréal, 7 to 18 July 2014

Montréal, 7 to 18 July 2014 INTERNATIONAL CIVIL AVIATION ORGANIZATION WORLD METEOROLOGICAL ORGANIZATION MET/14-WP/15 16/4/14 Meteorology (MET) Divisional Meeting (2014) Commission for Aeronautical Meteorology Fifteenth Session Montréal,