1. Retrieving the EUV solar spectrum from a selected set of lines for space weather purposes Jean Lilensten (LPG, Grenoble) Thierry Dudok de Wit (LPCE, Orléans) Jean Aboudarham (LESIA, Meudon) Pierre-Olivier Amblard (LIS, Grenoble) Frédéric Auchère (IAS, Orsay) Matthieu Kretzschmar (ROB, Brussels) + COST 724 team

2. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 2 INPUTS Specification of the ionosphere Ionosphere- thermosphere specification models Ionosphere- thermosphere specification models - electric fields - atmospheric tides - solar EUV flux USERS

3. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 3 Can one use proxies instead ? Dudok de Wit et al., 2007, accepted We define the degree of similarity between 2 observables by means of their Euclidian distance. The observables may be time series or measureable parameters such as densities …

4. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 4 In order to define the axis, we use the Singular Value Decomposition. In our case, 2 to 3 modes are enough to describe over 95% of the variance. Line or proxy 2 Line or proxy 1 Line or proxy 3 This degree is then represented on a dissimilarity map (called connnectivity map)

5. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 5 Sunspot number F10.7 index Mg II index wavelength [nm] They are all aligned !!! The proxie cannot and will never suffice to retrieve the solar spectrum whatever the combination (linear or not) Dudok et al., 2006

6. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 6 We need EUV measurements that are Spectrally resolved Calibrated Continuous (S. Solomon, NCAR)

7. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 7 Instead of measuring the full EUV spectrum, why not measure just a few spectral lines with a dedicated instrument, and use these as inputs ? 6 to 10 lines should be enough, based on the inversion of the DEM Kretzschmar et al., ASR 37 (2006) Our approach : use four years of EUV spectra from TIMED and a statistical approach to determine which spectral lines are the best for reconstructing the full spectrum (Dudok de Wit et al., Ann. Geoph (2005))

8. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 8 TIMED : 4 years of daily spectra (2002-2006) from the EGS/SEE spectrometer (level 2 data) spectral range : 26 - 195 nm spectral resolution : 0.4 nm flares are not (yet) included

9. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 9

10. 10 Our hypostheses Two lines with same dynamics Same underlying physics (?) No need to measure both lines simultaneously

11. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 11 How to choose the lines to be observed? Dendrogram of 38 spectral lines using an average distance linkage between all lines. Statistical analysis of TIMED/SEE data. Using two years of daily EUV spectra and classification techniques,

12. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 12 The EUV spectrum can indeed be retrieved from a small set of spectral lines, and with good accuracy (< 1% ave error). We have a rigorous statistical procedure for selecting those lines.

13. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 13 Which lines are the best ? The best choice is very much application dependent For aeronomy ? Best fit of the EUV irradiance ? Best fit of the variability ? … 10 best combinations of 6 lines Dudok de Wit et al., AnnGeo 2005

14. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 14 The « best » set depends on the application. An example : the ionosphere Lilensten et al., accepted in Ann. Geoph., 2007

15. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 15 Using a multidimensional scaling technique : H I at 102.572 nm CIII at 97.702 nm OV at 62.973 nm HeI at 58.433 nm FeXV at 28.415 nm HeII at 30.378 nm Allows to retrieve the full solar spectrum with a relative global error of 6.8 % and still fulfill ionospheric physics requirements.

16. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 16 Next future: To reduce the solar spectrum to a limited (3) set of characteristic spectra: very promissing method through positive source separation SEE (TIMED) : reconstruction < 2% Quiet sun contribution? Active zone contribution? Hot lines contribution?

17. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 17 Next future too: Use of a Neural Network (STSM T. Yapici and E. Altunas to France, scheduled in June and postponed after 1st September for budgetary reasons)

18. COST 724:Developing the scientific basis for monitoring, modelling and predicting Space Weather 18 Near future: LYRA, the Solar VUV radiometer on-board PROBA II J.-F. Hochedez et al., Adv. Space Res., 37, Iss 2, 303-312, 2006 http://lyra.oma.be/index.php 1/ Lyman-alpha (115-125 nm) 2/ the 200-220 nm range 3/ Al filter channels (17-70 nm 4/ MgF2 windows (120-220 nm), And EVE on-board SDO http://lasp.colorado.edu/eve/eve_home.html « unprecedented spectral resolution, temporal cadence, and precision. «