C3 -- Thermal Studies Meeting notes M. Weber & P. Boerner =========================================================== INDEX - Summary of main points from session - Rough notes of discussion on spectral modeling - Rough notes of discussion on inversion methods (blending into "DEM data products") - Rough notes of discussion on DEM data products =========================================================== Summary of session discussion: - Began with extended discussion of spectral models and associated uncertainties. - Concern about assumption of ionization equilibrium and recombination rates in the corona, especially in dynamic situations. Should be investigated further with MHD modeling. - Different abundance models can significantly affect thermal analysis. However, AIA channels are dominated by Fe stages, so there should be an investigation to verify that abundance is not an issue for AIA thermal analysis. - Systematic and calibration errors dominate uncertainties in DEM analysis. - Regina Soufli knows of high precision (order 1%) measurements of all of the AIA reflectivities vs wavelength. These should be collected, archived, and made available on an AIA website. - The discussion of AIA data products revolved around whether there should even *be* any pipelined DEM data products (eg, thermal maps), and how they could be fast, useful, and impervious to misuse. - To speed up production, and to avoid suggesting overly precise maps, there was a common idea to greatly reduce the resolution of these maps. Maybe do 5-6 megapixels per active region and per quiet Sun, or even just one mean DEM per active region. (This approach may imply feature recognition software.) - LOTS of concern whether end-users would understand the purpose of these data products as quick-look, and not for scientific analysis. How to communicate uncertainty ranges and emission measures? - Is there even a *demand* for these pipelined thermal maps? Maybe for space weather monitoring and for irradiance, but the usefulness for these have not been demonstrated. - Final thoughts from session organizers. - FOR GENERAL DEM ANALYSIS: - There are several issues to be resolved wrt spectral emission models and assumptions. These things can be investigated and addressed. A workshop with the database modelers and MHD modelers could accomplish a lot towards resolving these. - The interest in the issues associated with emission models and spectral codes was noticeably greater than the interest in the difficulties of DEM inversions. - FOR AIA THERMAL-MAP DATA-PRODUCTS, SPECIFICALLY: - There was a vocal minority that questions the value, cost, and pitfalls of providing pipelined AIA thermal maps. Here are some objections: - How to produce fast enough for 4kx4k images. - End-users will use these for science instead of just quick-look. - End-users will not understand how to interpret. May be overwhelmed, moreso if uncertainty and EM maps are accompanying. - Not demonstrated that there is any demand at all for these data products. - Organizer's opinion (MW): The value of pipelined data products is to provide a unique view on the data with which a scientist can discover interesting objects for further study. If there is a strong expectation that thermal maps will be useful in this way, then they should be provided. Thermal maps are arguably the Level-2 data products which best leverage the scientific strengths of AIA, but this needs to be weighed against the costs in developing and maintaining such a pipeline, and how it meets the scientific goals of the mission. (Furthermore, it might be possible to retroactively add this pipelined data product later after launch, if a demand for these evolves.) =========================================================== (Real-time notes) Issues of spectral modeling: - H. Peter: - Presentation to make the following points: - His group has done 3D MHD modeling of the TR and corona. - even in small boxed regions, the TR shows tremendous variability - C-IV and 304 good for seeing large variability in intensity and emission measure. - For relatively stationary situations, ionization timescales (tau_ion) are shorter than flow timescales (tau_flow). - BUT in flares or other dynamic situations, tau_ion > tau_flow (for TR). - EVEN EUV channels may be affected by this. - Studying a histogram of timescales from their MHD model, they find that for C-IV, tau_recombination > tau_flow, EVEN in less dynamic situations. This needs to be checked for Fe lines. - Based on these results, there is a concern whether DEM analysis (which assumes ionization balance) is even valid in dynamic corona. - Forward models (like theirs) may help analysis of validity limitations. - L. Culhane: Won't density be the most significant factor in the corona? - H.P.: One still needs to know if ionization equilibrium holds. - J. Schmelz: There are people looking at ionization non-equilibrium in the TR. - J. Cirtain: Wasn't there a paper with Mason that looked at low-corona loop models in ionization nonequilibrium, and they found that the densities were too low to agree with observations? - H.P: I think you're right, but these results should still be investigated. The timescales only differ by ~ factor of 2, not by orders of magnitude. - H. Warren: But shouldn't the problem be less in the corona? At flare energies and densities, there are FASTER reionization times. - M. Aschwanden: Don't we mostly observe COOLING loops in EUV channels? So reionization timescales shouldn't be an issue. - H.P.: Still have to worry about recombination. - J. Cirtain: Doppler shifts of plasma motion along the LOS may affect analysis of narrow spectral channels. - J. Lemen: Maybe C-IV and TR channels shouldn't be included in DEM analysis. What is the impact for selecting data products for AIA? - N. Nitta: - Presentation to make the following points: - Some time ago, M. Shimojo compared the Mewe and and different versions of the CHIANTI spectral models for the Yohkoh/SXT responses. During the Yohkoh mission, Mewe was the standard model. - Some lines went away and then returned over subsequent versions of CHIANTI. - APEC, CHINATI, and Mewe have converged for broad spectral ranges, BUT the ratio of the spectra from APEC and CHIANTI have noticeable differences over the range 160-200 Angstroms. - CHIANTI now has more lines than the other models. - J. Schmelz: Surprised there is still a difference between APEC and CHIANTI. - H. Winter: CHIANTI has drastically improved in the 1-50A range. - H. Warren: These codes don't account for the dependence of ionization fractions on density. This is incorporated in ADAS codes. - V. Kashyap: WRT APEC, it has a capability to handle uncertainties in wavelength and coefficients. - J. Schmelz: Has anyone looked at the effect of different spectral models in the AIA responses? - J. Lemen: How should these issues be addressed in the design of AIA data products, especially for an automated "pipeline" data product? - H. Winter: An important issue is how to propogate the errors from the spectral model details into the DEM analysis. - V. Kashyap: Systematic errors dominate. - J. Lemen: Also need to remember that AIA channel calibrations will evolve in the on-orbit environment. Can do some cross-calibration with EVE. - P. Boerner: There are plenty of pre-launch calibration issues, too. - V. Kashyap - Presentation to make the following points: - Shows AIA temperature response curves for various selections of abundances. - Selection of abundances makes a significant difference in the response curves, although the difference is mostly a magnification factor and doesn't affect the shapes of the curves much. - This effect may be manageable for narrowband instruments like AIA. - J. Schmelz: 6 of the AIA channels are dominated by Fe, so abundances issues are not a great issue. Also, it is hard to know which abundance model is best, but we know that the Feldman abundances are the worst. - V.K.: His MCMC method (Monte Carlo Markov Chain) for DEM analysis with high-resolution spectral data can include any number of uncertainties, even correlated uncertainties. Is it a reasonable idea to pre-compute responses for a limited set of ion balances? - P. Smith: In his experience, the people who calculate these spectral models are reluctant to provide uncertainties. In science, it should not be acceptable to provide numbers without uncertainties, so modellers should be persuaded to provide uncertainties with their models. - R. Soufli: We now have high precision (order 1%) measurements of the AIA refectivities vs wavelength. These should be collected. (General agreement from audience.) =========================================================== (Real-time notes) AIA DEM inversion methods (blending into DEM data products) - Mark = Mark Weber - Trae = Henry "Trae" Winter - Piet = Piet Martens - Jonathan = Jonathan Cirtain - Joan = Joan Schmelz - Markus = Markus Aschwanden - Harry = Harry Warren - Tom = Tom Woods - Joe = Joe Gurman - Jim = Jim Lemen - Peter = Peter Smith Mark: Introduction. Summary of topics and questions surrounding the different techniques used for DEM inversion (essentially contained on .ppt slide) Trae: How do you enforce non-negativity in direct inversion methods? Mark: Several possibilities: traditional regularization, SVD methods.Brief discussion of the SVD problem; is it solvable? Another possibility: Calculate a few (thousand) DEMs, and treat all observations as a case of one of these DEMs Piet: Re : the temperature gridding: why not just use 6 temperature bins for the six channels? Mark: The obvious solution isn't necessarily best; we could consider some non-uniform grid based on the temperature response functions. (Redirecting): Are thermal maps useful? Trae: "Temperature maps" -- basically EM-weighted T at each pixel -- might be useful. It's probably not necessary or possible to have full DEM maps at each pixel at high cadence. This brings up the question of computation time: e.g., Trae has a forward-fitting technique that runs much faster when it starts with a good guess. Is it possible to use a (fast) linear inversion technique to feed a fitting method? Johnathan: Is it really possible to produce even a quick-look T map? How much quicker is that than full DEM inversion? Joan: T maps can't/shouldn't really be a quick-look data product. For DEM inversion, combination of AIA with XRT should be very helpful for constraining high-temperature end of DEM Markus: T maps not useful for analysis of coronal structures because they cannot capture the details of overlapping loops etc. It might be useful to have T maps of the average, background corona. Harry: DEM data product could be useful as a bridge between AIA and EVE observations. In particular, spatially-resolved DEMs (spatially resolved by region, e.g. for an active region), could help in irradiance modeling. Tom: The EVE SAM could also be used to constrain the high end of the DEM (maybe). There are lots of instruments to combine. Joe: EIT observations have been used to produce T maps. In some ways, EIT observations are easier to invert for T maps. However, the reliability & usefulness of these T maps is not at all clear. Johnathan: What is the AIA data product? Is it sufficient for people to do their own temperature analysis? Jim: The level 1 data product is calibrated counts/s/pixel. Calibration will change during flight, but it is basically what you would need to do DEM inversion. Peter: Re: calibration, the thin-film filters have never really been measured; how good is the calibration on them? =========================================================== (Real-time notes) AIA DEM data products: - Most of this discussion focused on whether to make thermal maps with AIA data. - General discussion of using lower resolution, or regon-averaged values, to reduce complexity and computation time of thermal maps. - V. Kashyap: How about feature recognition? Could solve just 5-6 averaged pixels per active region. - M. Aschwanden: It would be useful to have a mean DEM per active region. - P. Martens: He is skeptical of thermal maps, as an AIA "pipelined" data product. He is worried that some end-users will take them and use them improperly for scientific analysis. (Metaphor of DEM data products as providing a gun for the end-user to shoot themself in the foot.) - J. Lemen: The easy answer is to not provide pipelined thermal data products. The slightly less easy answer to provide them with lots of warnings. - H. Peter: Maybe the thermal maps should be low resolution. - J. Wolfson: Raises question whether anyone strongly *wants* thermal map pipelined data products. - M. Aschwanden: People who want irradiance measurements may want pipelined thermal maps. Furthermore, thermal map data products should probably include emission measure maps and associated "error maps". Do visual maps work well for conveying this information? - D. McKenzie: Expresses concern that the problems of clearly presenting all this information may be not worth the effort. - Took an informal hands-raised survey: How many people think the difficulties and pitfalls of providing pipelined thermal map data products are not worth the trouble and/or resources? Result: Approximately 8 out of 20 audience members raised their hands. ===========================================================