Mem. S.A.It. Vol. 76, 941c© SAIt 2005 Memorie della

Image processing for theArcetri Solar Archive

M. Centrone, I. Ermolli and F. Giorgi

INAF Osservatorio Astronomico di Roma, via Frascati 33, 00040 Monte Porzio Catone,Italye-mail: name@mporzio.astro.it

Abstract. The modelling recently developed to reconstruct with high accuracy the mea-sured Total Solar Irradiance (TSI) variations, based on semi-empirical atmosphere modelsand observed distribution of the solar magnetic regions, can be applied to construct the TSIvariations back in time making use of observations stored on several historic photographicarchives. However, the analyses of images obtained through these archives is not a straight-forward task, because these images suffer of several defects originated by the acquisitiontechniques and the data storing. In this paper we summarize the processing applied to iden-tify solar features on the images obtained by the digitization of the Arcetri solar archive.

Key words. Sun: historical data – Sun: variability – Sun: magnetism

1. Introduction

A central question for understanding solar vari-ability is to what extent surface magnetism de-termines irradiance variations. Following ini-tial studies that used proxies of the magneticfield, more recent approaches to answering thisquestion use atmospheric models and observedsurface distributions of the solar magnetic fieldto reconstruct the irradiance. This approachhas succesfully reproduced with high accuracytotal and spectral irradiances measured fromspace, but has so far been restricted to MDI(Krivova et al. 2003), PSPT (Penza et al. 2003)and KPVT - MDI (Wenzler et al. 2005) data,available only for a period covering part of so-lar cycles 22 and 23. However, the temporalbaseline of the period analyzed through thismodelling can be remarcably increased takingadvantage of the great amount of highly valu-

Send offprint requests to: M. Centrone

able information stored in some existing solarphotographic archives. Those archive mainlycontain full disk observations performed at theCaII K, Hα and continuum radiation throughspectrographs. Thus, in analogy with the mod-elling applied to ”reconstruct” measured vari-ations, TSI variations can be extended backin time, assuming atmosphere models and ob-served distributions of the solar magnetic re-gions attained from the analysis of these pho-tographic archives. However, the analyses ofimages obtained by these archives is not astraightforward task. These images suffer ofseveral defects originated by the acquisitiontechniques, as well as by the data storing.In this paper we summarize the processing ap-plied to identify solar features on the imagesobtained by the digitization of the Arcetri solararchive.

942 Centrone et al.: Image processing for the Arcetri Solar Archive

2. The Arcetri Solar Archive

The Arcetri synoptic solar archive contains12917 plates of full-disk CaII K2,3 and Hα

spectroheliograms acquired during 5042 ob-serving days at the Arcetri G.B. Donati towerfrom 1925 to 1974 (Godoli e Righini 1950;Gasperini et al. 2004). A brief descriptionof the Arcetri instrument charactersistics atthe time of observations is given by Marcheiet al. (2005); Giorgi et al. (2005). Detailsabout the digitization work carried out by theCVS (Centro per lo Studio della Variabilitadel Sole) project at the Rome AstronomicalObservatory, with the sponsorship of RegioneLazio, can be found in Marchei et al. (2005).In brief, the archive digitization produced 16bit 8435×11153 pixel tiff images includingfour plates, from which 2k×2k 16 bit fits im-ages were singled out for each original plate.

3. Image processing

Prior to any processing the images are pre-analyzed and calibrated (Fig. 1) using dedi-cated procedures described by Giorgi et al.(2005). Some geometric and photometric is-sues have also to be taken into account beforethe application of the image decomposition forthe solar feature identification. In fact, work-ing with the data it became evident that someadditional processing would reduce the spuri-ous detections and enhance the feature iden-tification. In fact, specific defects can appearon a significant set of the digitized images. Forexample, the passage of clouds during obser-vations appear as a local darkening over thedisk, dust particles on the optical componentof the instrument result in dark lines across thedisk, modifications of the atmospheric turbu-lence produced a change in the image sharp-ness. The degradation effects produced by thefirst two issues can be solved through a properlimb darkening compensation, as described inthe following. Moreover, as the images cor-rupted by dust on the optics show lines in oneprodominant direction, this also makes it possi-ble to develop a simple method to detect and torestore them before the application of the limbdarkening compensation. At last, some pos-

sible instrumental defects also produced geo-metric distorsions on the image stored on theplates. The most frequent problem was due todifferencies on the velocity setting of the driv-ing motors which translate the input solar im-age and the photographic plate. The image re-sulted elliptic, the axes of the ellipse being par-allel and perpendicular to the slit. The differen-tial refraction of the atmosphere also resultedin an oblateness of the solar disk. Proceduresfor the recovery of geometric degradation ef-fects on Arcetri plates have not yet been devel-oped. These procedures have to take into ac-count the shape of the solar limb, that repre-sents the only way to infer accurately the ge-ometrical distorsions inside the disk. However,a first evaluation of the geometric distorsionsobtained by the ratio of vertical and orizontaldisk diameters showed that the ellipticity is oforder than 10% on the sample of most degra-dated images, while the median value for thewhole data collection resulted 3.75%.

4. Limb darkening compensation

Decomposition methods that allow the extrac-tion of the various disk features are usually ap-plied to images in which the solar limb dark-ening has been carefully removed. Moreover,the comparison of geometric and photomet-ric properties of the identified features on adaily basis also requires the normalization ofeach image of the data collection to the sameintensity scale. The assumption made in nor-malizing each image to the same intensityscale is that the quiet sun intensity does notchange with the solar cycle. Even though smallchanges of the quiet sun irradiance can notbe ruled out, these can probably only be de-tected through space based observations orground based measurements obtained with pre-cise photometric telescopes. It seems certainlyvery difficult a detection of these possible vari-ations through analyses of archive spectrohe-liograms. In fact, these observations even af-ter application of a careful photographic cal-ibration still contain appreciable variations ofthe disk average intensity due to variations ofthe sky transparency and of the exposure timeamong data collected on different days.

Centrone et al.: Image processing for the Arcetri Solar Archive 943

Fig. 1. Intensity profile along the marked row of the image, that has been pre-processed andphotographic calibrated as described by Giorgi et al. (2005).

Fig. 2. Intensity profile along the marked row of the image, that has been compensated for thelimb darkening variation as proposed by Brandt and Steinegger (1998).

Fig. 3. Intensity profile along the marked row of the image, that has been compensated for thelimb darkening variation through an iterative polynomical fitting applied by rows and by columnsas described in the text.

944 Centrone et al.: Image processing for the Arcetri Solar Archive

Several methods have been proposed in litera-ture (Johannesson et al. 1998; Worden, Whiteand Woods 1998; Caccin et al. 1998; Brandtand Steinegger 1998; Zharkova et al. 2003)to obtain a reliable limb darkening compen-sation on spectroheliograms. We show for ex-ample the results obtained by applying themethod proposed by Brandt and Steinegger

(1998) to the Arcetri data collection. Thismethod is based on the calculation of cumu-lative intensity histograms for rings of con-stant area, i.e. assuming a circularly symmet-ric fitting of the intensities on the solar disk.We found that this method allows limb dark-ening compensations and solar disk intensitycomputations that are very stable respect to theactivity level. However, the images obtainedthrough this method suffer for the presence oflarge scale non circularly-symmetric intensitypatterns (fig 2). Moreover, the limb darkeningis accurately compensated only in the about5% reduced inner part of the solar disk. Onaverage the limb darkening compensation isobtained through this method with an accu-racy worse than 10%. This accuracy can beimproved to better then 5% by dividing itera-tively each row and column of the photograph-ically calibrated image for a 2-nd order poly-nomial fit on the quiet solar background in-tensities. These intensities were obtained leav-ing out all the pixels with a contrast differ-ing more than 2σ respect to the values pre-liminary attained through a 4-th and 2-nd or-der polynomial fit performed by row and col-umn respectively. The application of a non-circularly symmetric fitting allows the com-pensation of the solar limb darkening togetherto the removal of large scale intensity pat-terns affecting the original spectroheliograms.In fact, by this method we make a rough cor-rection for the transparency fluctuations of theatmosphere during the observation, as well asfor the presence of scratches resulted by dust.However, we found that the images obtainedwith this iterative fitting also suffer for over-compensation effects (Fig. 3) around extendedactive regions that decrease the efficiency ofthe polynomial fitting. We found that the com-pensation accuracy increases remarcably byapplying the polynomical fitting already de-

scribed on the quiet sun background pixels se-lected by applying the method proposed byBrandt and Steinegger (1998). The contrastimages obtained in this way allow the appli-cation of simple feature identification methodsbased on intensity thresholding and revealedthemselves useful to perform analyses of mor-phological and photometric properties of activeregions on the solar disk.

5. Summary

This paper briefly describes the image pro-cessing developed for the analysis of the CaII K images obtained by the digitization ofthe Arcetri solar archive. The same processingis now being applied to Ca II K imagesobtained from the Mt Wilson and the Meudonarchives. The ability to perform a uniformprocessing of images coming from differenthistoric archives will allow a rather remarcableextension of the temporal baseline of the dataavailable for long term solar variability studies.

Acknowledgements. Bruno Caccin incited very of-ten the digitization of the Arcetri solar archive, con-scious of its wide scientific interest. We thank FabioCavallini and the Arcetri Solar Group for providingthe archive, thus allowing the fulfilment of the CVSproject, supported by Regione Lazio.

References

Brandt, P., & Steinegger, M. 1998, Solar Phys.,177, 287

Caccin, B., et al. 1998, Solar Phys., 177, 295Gasperini, A., et al. 2004, Giornale di

Astronomia, 3, 23Giorgi, F., et al. 2005, this volumeGodoli, G. and Righini, A., 1950, Mem. Soc.

Astron. Ital. 21, 4Krivova, N. A., et al. 2003, A&A, 399, L1Johannesson, A., et al. 1998, Solar Phys., 177,

265Marchei, E., et al. 2005, Mem. SAIt., in pressPenza, V., et al. 2003, ESA-SP 535, 299Wenzler, Th., et al. 2005, A&A, 432, 1057Worden, J.R., et al. 1998, Ap. J., 496, 998Zharkova, V. V., et al. 2003, Solar Phys., 214,

89