Prediction of solar flares on the basis of correlation with long-term irradiance and sunspot levels Ingolf E. Dammasch, Marie Dominique (ROB) SIDC Seminar, Brussels, 07 Mar 2014 LYRA the Large-Yield Radiometer onboard PROBA2

Contents LYRA, spectral response, data Poster for ESWW10 Development, future questions

LYRA: the Large-Yield RAdiometer 3 instrument units (redundancy) 4 spectral channels per head 3 types of detectors, Silicon + 2 types of diamond detectors (MSM, PIN): - radiation resistant - insensitive to visible light compared to Si detectors High cadence up to 100 Hz

Royal Observatory of Belgium (Brussels, B) Principal Investigator, overall design, onboard software specification, science operations PMOD/WRC (Davos, CH) Lead Co-Investigator, overall design and manufacturing Centre Spatial de Liège (B) Lead institute, project management, filters IMOMEC (Hasselt, B) Diamond detectors Max-Planck-Institut für Sonnensystemforschung (Lindau, D) calibration science Co-Is: BISA (Brussels, B), LPC2E (Orléans, F)… LYRA highlights

 4 spectral channels covering a wide emission temperature range  Redundancy (3 units) gathering three types of detectors  Rad-hard, solar-blind diamond UV sensors (PIN and MSM)  AXUV Si photodiodes  2 calibration LEDs per detector (λ = 465 nm and 390 nm)  High cadence (up to 100Hz)  Quasi-continuous acquisition during mission lifetime LyHzAlZr Unit1MSMPINMSMSi Unit2MSMPINMSM Unit3SiPINSi

SWAP and LYRA spectral intervals for solar flares, space weather, and aeronomy LYRA channel 1: the H I nm Lyman-alpha line ( nm) LYRA channel 2: the nm Herzberg continuum range (now nm) LYRA channel 3: the nm Aluminium filter range incl the He II 30.4 nm line (+ <5nm X-ray) LYRA channel 4: the 6-20 nm Zirconium filter range with highest solar variablility (+ <2nm X-ray) SWAP: the range around 17.4 nm including coronal lines like Fe IX and Fe X

LYRA spectral response

Spectral degradation after 200 days in space Experience from SOVA (1992/93) and LYRA (2010/11) combined (“molecular contamination on the first optical surface … caused by UV-induced polymerization”)

Reminder: LYRA spectral response channel 2-3: Aluminium filter, nominally 17-80nm channel 2-4: Zirconium filter, nominally 6-20nm additional SXR components <5 nm, <2 nm for comparison: GOES nm => Flares !

LYRA data product: 3day quicklook

LYRA data product: flare list

LYRA data product: GOES vs. LYRA proxies

LYRA data product: long-term solar levels

Contents LYRA, spectral response, data Poster for ESWW10 Development, future questions

“Level” Significant daily minimum, without flares or instrumental artefacts

“Variance” Daily minor-flaring activity, standard deviation in small corridor

“Level” 100 values (*) closest around LYRA ch2-4 selected from 1300 observations => estimated distribution of flare strengths Same for LYRA ch2-3, GOES, DSSN => forecast based on 400 values

“Variance” 100 values (*) closest around LYRA ch2-4 selected from 1300 observations => estimated distribution of flare strengths Same for LYRA ch2-3, GOES, DSSN => forecast based on 400 values

“Level” – daily forecast

“Variance” – daily forecast

Warnings This is a statistical flare forecast (“Bayesian approach”). M- and X-flares are so exceptional that the estimated median will always stay below. Probabilities may rise from 0% to 30-40% (M) or 5-10% (X). It is not assumed that statistics like these can substitute a space weather forecaster's experience. Magnetic structures are not taken into consideration. But statements like the following become possible: “When the GOES level rises to B7, one has an almost 50% chance of observing an M-flare.” “No X-flare ever occurred while LYRA ch2-3 was below W/m², or LYRA ch2-4 was below W/m².”

“Level” Test Aug-Oct 2013 Method changes slower Median leads to underestimation during high activity Probabilities reflect situation better

“Variance” Test Aug-Oct 2013 Method follows closer Median leads to underestimation during high activity Probabilities reflect situation better

Contents LYRA, spectral response, data Poster for ESWW10 Development, future questions

Three months later… “Does it make sense?” – “Yes”, said Mike Wheatland (Univ. Sidney, invited lecture at ESWW10) Second activity peak of cycle 24 – does it change the statistics? How to evaluate a forecast which consists of more than one value? Are my methods better than the “Yesterday’s Weather” hypothesis? How can they be improved? Which forecasting parameter is the most reliable? Are our space weather forecasters interested?

Still problems with under- estimations in periods of high activity

Skill scores Create bins of probabilities of certain events (M-flare prediction between 0-5%, 5-10%, etc) and check the real percentage of events in these days (“reliability plot”) Per month or per week, check the prediction of events like M- days (=sum of M-flare probabilities) against the realized number of events (“prediction of event days”) Six months of prediction data exist, calculations TBD

Results to be presented An abstract was submitted to COSPAR Session D2.2-E3.2 “Space Climate” Title “Long-term irradiance observation and short-term flare prediction with LYRA on PROBA2”

Please visit