Mem. S.A.It. Vol. 76, 862c© SAIt 2005 Memorie della

The solar photograph archive of theMount Wilson Observatory

A resource for a century of digital data

S. Lefebvre, R. K. Ulrich, L. S. Webster, F. Varadi, J. Javaraiah, L. Bertello,L. Werden, J. E. Boyden and P. Gilman

University of California at Los Angeles, Department of Physics and Astronomy, Physicsand Astronomy Building, 430 Portola Plaza, Box 951547, Los Angeles, CA 90095-1547,USA e-mail: ulrich@astro.ucla.edu

Abstract. The solar telescopes and spectroheliographs of the Mount Wilson Observatorywere among the earliest modern facilities for the study of the solar surface. The photo-graphic collection of the solar program at Mt. Wilson begins in 1894 and continues to thepresent day. A program to digitize and distribute the images in this collection was begunat UCLA in 2003 and is now making available the first of the catalogued and catago-rized images from the CaK sequence. Most of the instrumentation with which the im-ages were obtained is still available although in a disassembled form. Original log bookshave been digitized and associated with the images so that a maximum of scientific re-turn can be obtained from the data base. The present range of images available fromwww.astro.ucla.edu/ ∼ulrich extends from late 1915 to 1952. Each image has been dig-itized with 12-bit precision and represented in a 16-bit format. These images are each 13Mbytes in size and larger than will be the final product images since not all image de-fects have been mitigated at this time. The radii and centers of the solar images have beendetermined and are included in the available data files. Optical vignetting by the systemintroduces an intensity gradient of known magnitude that can be used to help characterizethe photograph plates. The roll angle of the images has yet to be determined.

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

1. Introduction

The Mt. Wilson Solar Photographic ArchiveDIgitization Project (Mt. Wilson SPADIP)makes available to the scientific communityin digital form solar images in the archivesof the Carnegie Observatories. These imagesdate from 1894 and include many which can

Send offprint requests to: R. Ulrich

be judged to be of superb quality by modernstandards. These images will permit a varietyof retrospective analysis of the state of solarmagnetism and provide a temporal baseline ofabout 100 years for many solar properties. The20th century was a period of increased anthro-pogenic production of greenhouse gases whichmight contribute to global climate change. TheSun could also be a factor in global climate

Lefebvre et al.: Mt. Wilson solar image archive 863

change and the data provide by this digitizationwill allow the scientific community to freelyexamine some properties of solar magnetismover the 20th century. The objectives of theproject are:

1. to preserve the original plate material fromtime degradation;

2. to provide a web-based digital archive withincorporated log book parameters;

3. and to calibrate the images as much as pos-sible.

This digitized archive should permit the studyof a wide range of topics including many thatwe cannot anticipate. The distribution of theimages is unrestricted by means of a websiteand enable community utilization of this his-torical database. In addition, we have identifiedthree specific problems that can be addressedwith this database. We will carry out studies ofthe following questions:

1. How is the strength and distribution of theSun’s weak magnetic field related to thesunspot number?

2. What is the temporal history of the Sun’ssurface differential rotation?

3. What is the history of the strength anddistribution sunspot darkening and facularbrightening?

2. Type of data and management

2.1. History of the archive

Mt. Wilson solar Observatory was founded in1904 by George Ellery Hale under the auspicesof the Carnegie Institution of Washington.In that year, Hale brought the Snow SolarTelescope from Yerkes Observatory in south-ern Wisconsin to the sunnier and steadier skiesof Mt. Wilson to continue his studies of theSun. The 60-foot tower was the first tower con-figuration on Mt. Wilson. Hale used this tele-scope to identify magnetic fields in sunspotsfor the firs time.The program of Calcium spec-troheliograms began prior to Hale’s selectionof Mt. Wilson for the observatory. The ear-liest plates were obtained at the KenwoodObservatory during a period between 1891 and

1895. The earliest plates are of variable qual-ity. Beginning in 1905, several series of cal-cium spetroheliograms, whose A-series, wereobtaines at the 60-foot tower telescope and atthe Snow Telecope. And in 1905, broad-bandimages called white light directs are included.Digitized forms of these images are includedin the digital Mt. Wilson solar photographicarchive.

2.2. The plates and scans

The archive contains over 150,000 images ofthe Sun which were acquired over a timespan in excess of 100 years. The glass andacetate negatives are stored and maintainedat the Pasadena California offices of theObservatories of the Carnegie Institution ofWashington. The images include over 43,000broad-band images called White Light Directs,over 35,000 CaK spectroheliograms and over74,000 Hα spectroheliograms. Figure 1 showsthe coverage of the Calcium spectroheliogramsfrom 1915 to 1952. Before 1940, there is anaverage of 400 images per year, and 700 after1940.

Fig. 1. Number of CaK spectroheliograms per year.

The plates are fragile and must be han-dled with care in order to preserve their qual-ity. Prior envelopes, where plates were stored,have glue seams and higher acid content, thewooden cases outgas harmful chemicals. Now,plates are in archival 4-flap housing in inertboxes without glue. The photographic plateshave been scanned in groups of three with eachplate having up to 4 images. Most early plates

864 Lefebvre et al.: Mt. Wilson solar image archive

have two pairs with each pair being exposedsimultaneously. One image of the pair is at thecenter of the CaK line and includes the light ofCaK2V, K3 and K2R. The second image of thepair is in the light of the nearby part of the K1line and does not usually include other spectralfeatures. We do not use the data from the sec-ond images although they are included in thegeneral scans. The scans are being carried outusing an Eskographics F14 scanner. The solarimages have radii of about 1000 pixels and thedigitization provides about 12 bits of signifi-cant data. The scanner output images are in a16-bit TIF format with the intensity values be-ing stored as positive integers. We refer to theseinitial scans as fulldeck scans since all imagesare scanned simultaneously using the full deckof the scanner.

2.3. Log book information and scanidentification

Data from the original log books has beenentered manually and examined for validity.These digitized log files are downloaded to anarchive database. The solar images on the fulldeck scans are then tagged with correspond-ing digital log book entries by visually examin-ing each image on the scan for written markinggiving the date, time and observation ID num-ber. This tagging step links the time and dateof observation to the image position on the fulldeck scan. Using the tag information, an ex-traction is done with software written specifi-cally for this purpose. Plate geometry and so-lar image geometry are essential in this step.Each extracted image is stored as a separate16-bit TIF file and the information from thedatabase is recorded as the figure caption of theimage in a format that can be used during sub-sequent processing steps. This initial extractionprovides the first indication of the location ofthe solar image since the extraction softwarefinds a circular structure in the expected posi-tion and then picks a 2601 by 2601 pixel arraywith this circle of roughly 1000 pixels radius atits center.

2.4. The website

A web site has been developpedand is always updated. The im-ages are freely available throughhttp://www.astro.ucla.edu/∼ulrich/MW SPADP/.Raw images as well as images subjectedto partial processing are included. On thiswebsite, browseable images with thumbnailsare presently available from 1915 to 1952 andpermit easily to the user to choose the imageof interest for him. An example of thumbnailsis presented on Figure 2.

Fig. 2. Example of thumbnail issued from theSPADIP website.

Reduced-size science quality FITS fileswith image center as well as vertical and hor-izontal radii determined are also available forthe same years. Summary listings of the radiussolution results are found in each year’s direc-tory. Fits file description with a brief definitionof the algorithms for finding the radii and im-age centers is provided. A listing of the quan-tities provided in the summary files in eachyearly directory is accessible. The precised al-gorithms used are in FORTRAN and is also onthe website.

Figure 3 shows an example of Calciumspectroheliogram taken on June 25th 1928.Notice the good quality of this negative image:

Lefebvre et al.: Mt. Wilson solar image archive 865

a lot of plages anf faculae can be seen even athigher latitude.

Fig. 3. Example of Calcium spectroheliogram takenon June 25th 1928 (negative image).

3. Calibration of the image

A procedure of calibration has been devel-opped. We give here a summary and the detailscan be found on the web site.

1. Dust and pit removal: these images presentnotably some dust and pit which is im-portant to remove (or at least reduce). Todo this, a laplacian filter is applied to theimage to locate discordant groups whosescale is smaller than the spatial resolutionof the observing system. Typically 4,000 to100,000 pixels out of 6.7 Mpix have theirvalues replaced in this process.

2. Initial guess of image center: The solar im-age is identified in the 2601 by 2601 pixelarray by first calculating the rms deviationof square sub-arrays of 5 by 5 pixels. Usingthe average of all these rms values as a di-viding point, the centroid of all pixels hav-ing larger than average rms values is takenas the center of the solar image.

3. Image geometry: A strategy was adoptedwhereby the search in the image is forradii and image centers associated with

the x (scan/dipersion direction) and y(cross-dispersion/slit direction) indepen-dently. Four separate locations are sought:x+, x-, y+ and y-. The center point in eachdirection is then the average of the + and- limb positions along each of these axes.With a tentative center, four quadrants ofthe image correspond to each of these limbpositions. The expected time dependencevalue of the radius is used to constrain thesolutions for the radii within bounds thatare tightened successively during an itera-tive process.

4. Intensity gradient determination: The nextstep in finding the solar image limb isto calculate a gradient image using thefirst guess image center. Averages are cal-cultated for two 5 by 7 blocks of pixelswhich are symmetrically centered on thetarget pixel. The inner block average is sub-stracted from the outer block average toform a smoothed measure of the intensityslope at each point.

5. Reference gradient: Due to exposure varia-tions and optical system vignetting, the in-tensity and its gradient are highly variablefrom one image to another. In order to usethe gradient images in spite of these vari-ations, the first step is to determine typicalvalues of the gradient for each quadrant byaveraging the absolute value of the gradi-ent in a strip of width 20 percent of the ar-ray dimension centered on the axes whichgo through the trial image center.

6. Circular arc averaging - edge definition:Still using the trial image center, a series ofcircular arcs can be defined as a function oftrial image radii in each quadrant. The se-ries of trial radii begins at a distance largerthan the expected solar radius and is suc-cessively decreased until the gradient av-eraged along this arc segment exceeds 20percent of the reference gradient.

7. Scattered light correction: Many of the im-ages have scattered light primarily due tocondition of the dispersion grating whichwas in a pit below the observing floorand subject to conditions of variable andhigh humidity. During each sequence ofdecreasing trial radius, the average of the

866 Lefebvre et al.: Mt. Wilson solar image archive

Fig. 4. Example of center-to-limb compensation for two images. Top: image taken on August 1st 1937;bottom: image taken on May 2nd 1939. From left to right: original image, extracted background image andflattened image. The roll angles have not been set to anything in particular and the pole position markers isindicated on each image.

gradients for the trial values more than twosteps previously and less than 10 steps pre-viously is calculated. This average is takento be representative of the scattered lightand is substracted from the gradient at thetrial radius under consideration. The scat-tered light correction is limited to 50 per-cent of the reference.

8. Veto conditions: In addition to the basic cri-terion above, a number of veto conditionsare tested and the trial radii will be de-creased further if any of the veto conditionsare found to be true. The veto conditionsare: plate scratches produce paired and op-posing gradients and plate edges producestrong gradients, which are either horizon-tal or vertical and which are not curved. Ifthe gradient maximum is found to have oneof these properties, it is identified as eithera line (paired positive and negative) or anedge (unpaired) and is vetoed. Very strong

dust spots or plate pits produce very high,paired slopes of opposite sign. If a dust spotis found along the averaging arc, the gra-dient maximum is vetoed. An area off theplate has a constant value equal to the clearscan result so that its rms is zero. Such ar-eas are identified as blank and if the radiusscan stops at a boundary of a blank area,it is vetoed. Finally, if the trial radius be-comes smaller than 95 percent of the ex-pected radius, the search is stopped. Afterthe search is stopped, the test causing theprevious point to be unacceptable is ex-amined and its nature is used to calculatean image demerit value. The proper test tohave failed is that the prior gradient was toosmall.

9. Image center iteration: After the radii aredetermined in the four quadrants, a newtrial center is taken having its x and y coor-dinates as the averages of (x+ and x-) and

Lefebvre et al.: Mt. Wilson solar image archive 867

(y+ and y-) respectively. The quadrant ra-dius searches are repeated iteratively untilthe trial center no longer changes.

10. Image demerits: After the sequence con-verges, a demerit value is computed basedon veto conditions and the deviation of theimage from its expected size and its devia-tion from being circular.

11. Center-to-limb variation compensation:One of the main problem we encouteredduring the calibration process is the pres-ence of a vignetting function. This functionis linked to the centering of the optical sys-tem and more precisely to the relative posi-tion between the pupil and the grating. Dueto that, the intensity and its gradient arehigly variable from one image to another.Of course, it is important to compensatefor the center-to-limb variation mainly dueto the presence of this vignetting function.Several methods can be used. Here we usean iterative procedure with an adaptativemedian filter: we extract the backgroundand divide the image by the background toobtain a flat image where the distributionof spots and faculae can be tracked to haveaccess, for example, to the differential rota-tion rate. Figure 4 shows two examples ofextraction of the background.

Other steps in the calibration, as the roll an-gle determination, are under investigation.

4. Future studies at UCLA

4.1. Solar differential rotation rate

The chromospheric network is clearly visibleon a substantial fraction of the archive im-ages from the CaK spectroheliogram sequence.Snodgrass and Ulrich (1990) have derived a ro-tation curve from the synoptic magnetogramsequence of the 150-foot tower. The CaK spec-troheliogram sequence can be used in a sim-ilar fashion to provide a history of this pro-cess extending back to the 1915 era. If the pat-tern of differential rotation rate is in fact con-nected with the overall strength of the sunspotcycle, we should be able to detect this vari-ability through the cross-correlation methods

of Snodgrass and Ulrich (1990) applied tothe CaK spetroheliograms sequence. Since thestrength of the solar cycle has increased overthe 20th century, any associated change in therotation rate should be detectable.

4.2. Weak field magnetic fields

The strength of the widely distributed weaksolar magnetic fields will be studied throughtexamination of the brightness of the CaK lineemission over regions of the solar surface notdirectly associated with sunspots.

4.3. Total solar irradiance reconstruction

In collaboration with J. Pap from the GoddardEarth Sciences and Technology Center of theUniversity of Maryland, the time dependenceof sunspot and plage areas using the broadbandimages will be determined and the influence ofthese variations on the Total Solar Irradiance(TSI) will be estimated in order to reconstructan improved history of the solar output energy.

5. Conclusions

As of June 2005, more than 20,000 CaK im-ages spanning the period 1915 to 1952 havebeen scanned, identified with log book en-tries, analyzed for radius and image centerand placed on the project web page as fitsfiles. Vignetting and plate sensitivity func-tions are under investigation. Image roll an-gle needs to be determined and scattered lightneeds to be corrected. An analogous programwith Arcetri solar archive is under progress atRome Observatory under the responsablility ofI. Ermolli and her team (Marchei et al. 2005).

Acknowledgements. This project is sponsored byNSF-ATM, AST and NASA-LWS and supported bythe solar physics group at Stanford university.

References

Snodgrass, H. B. & Ulrich, R. K. 1990, ApJ,351, 309

Marchei, E., et al. 2005, this volume