Glenn Schneider (Steward Observatory, University of Arizona)
in collaboration with Joel Moskowitz, M.D.

For my 28th total solar eclipse (prior chronology here), I joined Rick Brown's group of eclipse chasers, observing from the grounds of the Wuhan Bioengineering Institute (WBI), located to the north-east of the city of Wuhan, China.  TSE2009 offered the possibility of an extraordinarily long duration of 6m 41s from a ship (the Cosa Classica, moving at 11 knots in the "right" direction at the point of maximum eclipse in the Pacific Ocean). However, I had had enough of sea- (and aircraft-) going eclipses for a while and wanted my feet firmly on terra firma, and the firmer the better!  Hence the appeal of the large, open area, concrete quad at WBI also offering good horizon views for the approaching and receding lunar shadow. With cooperative weather, an ideal location.  A few pictures of the WBI and the eclipse observing site on its campus are shown below.

For TSE2009, Joel Moskowitz and I with Rick Brown's eclipse chaser's group set up on the WBI quad, right at the location of the "dark stain" at the right of the bottom photograph (taken a year before TSE2009 by Rick Brown, looking east toward the direction of the to-be-eclipsed Sun; north is to the left) is at the GPS measured (WGS84) geographical coordinates noted above.

From this location we had anticipated, and planned for, 5m 29.6 of totality predicated upon the predicted local eclipse circumstances as follows:

Wuhan Bioengineering Institute Eclipse Site
Local Eclipse Circumstances

       C1:   00h 15m 10.2s   +32.6°
       C2:   01h 24m 14.4s   +47.5°
       Mid:  01h 26m 59.2s   +48.1°
       C3:   01h 29m 44.0s   +48.7°
       C4:   02h 46m 49.3s   +64.9°

       Penumbral Duration: 2h 31m 39.1s
       Umbra Duration:         5m 29.6s
       Magnitude: 1.035

  Lunar Limb Profile Corrections Applied
  Atmospheric Refraction Corrections Applied
 Ephemerides: LE/DE 406
 Delta-T = +66.448s
 C.O.M to C.O.F Offset = (+0.50”,-0.25”)

But, we saw "only" approximately 94 seconds. I say "only", in quotes, because this alone was still very significantly longer than five of the (27) earlier times I have been enveloped by the lunar umbra: (1977 [38s], 1986 [0 to 5 seconds depending upon how defined and measured], 1995 [57s], 2002 [26s], and 2005 [32s]).  Indeed, it is a sobering thought that the amount of totality we saw of TSE2009 was essentially equal to that experienced in total from 4 of those 5 aforementioned eclipses!  The disparity between the ~ 5m 30s of totality predicted and the ~ 94 s we saw, however, was not because our eclipse predictions were off the mark (they were not!), but rather for this reason as so well captured in Chinese newspapers the day after the eclipse:

But note, despite the (perhaps somewhat over-embellished) reactions of the cartoon observers being initially much like our own, the eclipsed Sun for us was still partially visible during totality. And thus, for us at WBI, TSE2009 became the totality for which for which  the glass was 1/4 full rather than 3/4 empty.  From our site at the WBI, happily, some coronal photons did make it through the shifting clouds, contiguously for the last minute and a half plus of totality (see below), thus enabling a partial victory snatched from the jaws of a cloudy defeat.

The day before the eclipse we had assessed the weather situation with other like-minded and weather-causality savvy umbraphiles Jay Friedland and Roland Burley while at the Wuhan airport -- with in part an advanced option in mind to relocate eastward to Shanghai, or westward to Chongquin, by commercial air if needed to gain an advantage.  (It was particularly welcome to have Roland, a 747 pilot for Cathey Pacific Airlines who lives in China, with first-hand insights into the local weather in the loop - thanks Roland). But to all of us, it was readily clear (or, I should say apparent) that Shanghai and environs was best avoided and we anticipated (at least some) eclipse-chasers relocating in our direction.  There did not seem to be any unambiguous advantage for Chongquin over Wuhan or vice versa.  So,  we decided to stay put for that moment but keep a continuous eye on the weather with later local relocation possibilities using the chartered busses Rick had arranged for. 

"Last minute" dashes to out run fronts or find holes in cloud cover (as I have before done when meaningfully informed, e.g., TSE1976TSE1979 and TSE1997) did not apply. Over the entire Wuhan region of short time-scale relocatability, the local cloud conditions after sunrise were spatially variable and quite broken on small scales as noted in situ and with satellite imagery.  Indicators of potential advantage with relocation were not strong (that perhaps an understatement) and a decision for pro-active relocation was not well constrained. Without the benefit of post-eclipse hindsight, given all a priori indicators available, such action could easily have made things worse, rather than better.  It was, in the end, despite all informed eyes on the weather, significantly the luck of the draw.   Solar visibility varied on distance scales of kilometers or less and timescales of at most minutes (and often seconds).

In the prophetic words of Mel Brook's "Twelve Chairs" theme, though day-after informed history might (unreasonably) suggest otherwise, as totality approached we did "hope for the best, expect the worst".  And, for better or worse, we were rewarded with less than complete success, by no means suffered a complete failure.

My autonomous photographic plans for totality were centered on using the same 1.2 meter EFL f/12, coelestat-fed, eclipse camera (under UMBRAPHILE control), aka my "lug-a-scope", that I had used several times before (see below) in only slightly different incarnations, e.g. recently for TSE 2006, 2002, and 2001.  The "plan" was to use, for the penultimate time in the remainder of my eclipse chasing life, my absolute favorite film for TSEs: Kodachrome 25P (sadly now, like all other now discontinued Kodachrome films, as prophetically lamented by Paul Simon*, destined for extinction in Dec. 2010).  The presence of partially obscuring (but often not fully opaque) cloud during the partial phase of eclipse ingress, made me switch out to a somewhat faster Kodachrome 64. But, in the end, still not fast enough to meaningfully capture totality as the long(est) exposures were pre-programmed to be taken at mid-eclipse, a time at which the clouds were fully veiling the view of totality.

The 1.2 meter EFL eclipse camera (4" objective lens obscured) pointed at the tracking coelestat and aligned on the Sun about half an hour before totality.  Note the presence, then, of only diffuse and  low-contrast shadows -- the clouds at that instant acting as a nearly perfect optical density 5 filter for the partial eclipse in progress!  Joel Moskowitz (red shirt), my daughter Maia (yellow shirt) and myself (checking computer operations) all hopefully adorned with dominant-eye eyepatches for dark adaption.

UNFORTUNATELY... The first 4 minutes of totality were completely obscured by clouds!  A bit past mid-eclipse I sat down on the ground still with a small glimmer of hope, but nonetheless feeling dejected (my daughter says I was crying, but I stoically maintain it was just sweat dripping from the heat and humidity; sniff, sniff...).  But then, about 94 seconds before C3 there was a tiny brightening in the sky and... a small "hole" and coronal photons were pouring through into our eyeballs, with increasing brightness over time - with astonished shouts of "THERE IT IS", etc.  Actually, I should specifically say INNER coronal photons, as the remaining clouds were dense enough (unfortunately) to obscure the mid and outer corona.  But over the next minute and a half -- with thin wafting clouds -- we saw totality!  Not the "clear sky" totality we had wished for, but an ethereal totality nonetheless.  The inner corona was viewed for about a minute and a half through those sparse and flowing clouds, followed by a beautiful chromospheric arc just before third contact, and a glorious 3rd contact diamond ring. Below is a wide-field photo taken by Geoff Simms who was set up very close to me that illustrates the situation soon after the eclipsed Sun partially broke through the rapidly moving clouds. 

The broken wafting clouds over WBI gave us partially obscured, but contiguous, views of the totality in the last minute and a half before third contact, as captured in this dramatic photograph by Geoff Simms (click on the picture to see at a larger image scale and higher resolution).  Note the chromospheric light illuminating the clouds toward the ESE (right).  Also note the "hole" in the clouds to the south (right of the Sun).  If only we had been shifted a couple of kilometers over...

When the Clouds Parted  —  Second-by-Second

Click on image mosic above to see a larger version.
How much totality did we see?  That is somewhat subjective. The lunar limb (or at least a part of it) was clearly seen in silhouette against the corona at least 94 seconds before 3rd contact - which is my estimation,  but I'll let you decide.  The above matrix of 1 second cadence images of totality (time increasing top left to lower right, nine seconds per row, north to the right in each frame) begins 100 seconds before third contact (bottom left image at  01:29:44 UT with the onset of the 3rd contact diamond ring).

The density of the clouds was incompatible (as anticipated) with the pre-programmed exposure sequence I had planned for my lug-a-scope (with relatively low speed film and slow f/ ratio) – so, unfortunately, no pictures of any use from that high resolution camera. But, Joel Moskowitz kept his HD video camera (see below) running and adjusted to a higher gain and sensitivity (with readout noise penalty -- but no choice!) and captured what we saw (though not quite as the eye perceived it).

Click  HERE to see a low-resolution, but full frame, copy of Joel's HD video.

You can see a full-resolution version of the HD movie, but with just the region (of interest) around the Sun extracted, by clicking on the (Quicktime) movie icon to the right (in yellow frame). WARNING: This is a 709 Mbyte file (long download time, but worth waiting for).  Here, I have done some global brightness and contrast stretching (same to all frames; x1.35 brightness, x2 contrast) to modestly enhance coronal visibility in the presence of obscuring clouds.  The video begins a few seconds before the corona first became faintly evident. From about 10  to 25 seconds into the video the image becomes "shaky" as Joel manually adjusted the camera's aperture (16 seconds into the video a sudden change in brightness is seen as a result).
and contrast



The above HD video very nicely captures the streaming nature of the foreground clouds (as David Makepeace called similar clouds in Moganshan: "Misty") - but due to the dominant low light levels, the signal-to-nose is low, and the individual frames are relatively "noisy". 

TSE2009 "Movie" after Image Processing

I have post-processed the appx 3600 raw frames extracted from the high-res HD video (above), to create a higher SNR,"time-lapsed" TSE2009 totality movie (replayed in real time, with an effective processed  inter-frame cadence of 1 second).  Click on the movie icon to the right (in green frame) to view the image processed movie (a 139 Mbyte file) at a real-time cadence and full resolution. Details of the image processing are given below. 



Sound on.

The individual post-processed frames contained in the high-resolution post-processed time-lapse movie, are shown, at low-resolution in the mosaic to the left. Time proceeds from upper left to lower right. The presence of partially obscuring cloud is quite obvious with the advent of the third contact diamond ring (last row).  Click on the image mosaic to see at higher, but not full (as in the movie), resolution.

N.B. In all three movies, north is  toward the right.

Need a QuickTime viewer?  Click HERE.


By further combining selected post-processed images (e.g., in the above example using  16 post processed frames from images 60-75 as shown in the image mosaic), to some degree the rapidly moving clouds can be partially filtered away -- better revealing some details of the true  inner coronal structure.  Additionally in this image, rotated approximately to the orientation of the Sun as seen from Wuhan (same as the three single frame images at the top of the page), a (30 degree rotational kernel) low-amplitude azimuthal spatial fiter has been applied to modestly enhance the visibilty of inner coronal features.

The HD video of  totality was acquired by Joel Moskowitz using a SONY PMW-EX1 HD camcorder with a Century Optics 1.6x teleconverter.  The HD camcorder was mounted on a Nikon DX3 DSLR equipped with a Takahashi Sky 90.  Both cameras were co-mounted on an Astrotrac equatorial mount with image focal plane detectors oriented to place the anticipated long axis of the corona along the long axes of the sensors. (The automated DX3 imaging program failed, like the "lug-a-scope" program, due to image underexposure arising from variable cloud cover incompatible with the pre-programmed UMBRAPHILE imaging sequence).

Summary Details of Video Image Post-Processing (for those who want to know):

While the clouds of Wuhan, of course, remain the limiting factor in coronal visibility, with frame-by-frame and inter-image post-processing of the HD video, I have been able to:

  (a) increase the photon SNR in the low-flux regime,
  (b) largely mitigate (reduce) noise due to the camera readout electronics (and/or digital encoding),  
  (c) improve image contrast,
  (d) compensate (or largely so) for sensitivity and color balance variations with changing cloud opacity,
  (e) correct for jitter and image wander when the camera was being adjusted.

Of course, a basic tenant of information processing (as well as many other things in the physical world) may be summarized as “there is no free lunch”. So, all of the above “improvements” (and I’ll let you judge if they are) have been done at the expense of temporal sampling -- that is the primary trade space.  With some other trades, to simultaneously realize the above goals, I found (experimentally) it necessary to make use of the information in approximately 30 (contiguous) video frames (spanning 1 second in time given the frame rate) to create one “output frame” representative of a one second, (nominally) 30 image, SNR weighted median-combined "snapshot".  Thus, the processed “movie” has a cadence of (only) 1 fps, and (thus) is played as packaged as a QuickTime movie at that rate -- appropriate to show what we saw of totality in “real time”.  I.e., it “looks like” a time-lapsed movie (or as one would get from a photographic intervelometer sequence).  The original HD movie flows much more “smoothly” in time. But the processed movie, while much more temporally discrete, is composed of post-processed frames that are a significant improvement compared to the individual input HD video frames (see below). 

A representative example of the improvement in image quality acheived is illustrated by comparing a raw (but 1.35x brightness, 2x contrast renormalized) HD frame (left, from 1m 32s into the video, just as the chromospheric arc begins to appear) with a corresponding frame post-processed (right) as described below.

The original movie with a 30 fps cadence, of course, “looks” more natural in terms of not being a series of discrete pictures and the "wafting" of the (#*@^&!) misty-clouds is much more "esthetically pleasing" (in a very ironic way, as we wish they were not there!). Download the Post-Processed movie from here (same as link above in the green frame; note: this is a 136 Mbyte QuickTime movie):


Below, I leave out processing details of minutia, but note the process was a bit painful (or at least time consuming) because: (a) it was not fully automated in some steps and (b) also due to the low SNR of the input frames I had to manually inspected  to reject outliers from all appx 3600 frames individually.  In all, only a small number of input frames were rejected.  Most of those were near the beginning of the raw video, where the cloud density was still extremely high and during times where Joel was manually handling the camera (leading to image smear within individual exposures, and changing inter-frame camera sensitivities).

1. Extract image sub-arrays from a fixed region of the HD field containing the Sun for every video frame – i.e., create 3600 raw image files (of just the region of interest) for subsequent processing.

2. Examine ("by eye") every individual raw color frame and reject those that had significant image smear (due to camera handling) or other artifacts.  (Only a very small number were rejected, but importantly so, so as not to bias later processing).

3. Color separate R, G, B planes of each image into separate files (so 3 x 3600 files input files to process).

4. Quantitatively determine image mis-registrations. Adaptively cross-correlate sequential frames (separately for R, G, B) to determine position offsets from image jitter.  Combine cross-correlation results (with SNR weighting) in each color to determine and correct for image mis-registration (jitter and wander),  cross-checked chi-square minimized image first-differences.  This is a variant of the process I had developed originally for "dejittering" Jay Friedland's TSE2003 hand-held video.  For these interested in details SEE HERE.

5. Correct for image jitter and tracking errors.  Register (in X/Y translation only, not rotation) all images using (4) to a common reference.  Translate via resembling each R, G, B, image (separately) onto a same-sized output image grid with sinc-apodized bi-cubic interpolation. (Detail: This (and also step 7) were done using an IDL-based S/W application package called "idp3", developed by the NICMOS (Near Infrared Camera and Multi-Object Spectrometer) team for Hubble Space Telescope image processing - but is widely applicable to other astronomical image processing). [A couple of papers of possible interest: 1 (see section 3), 2]

6. Bin into 1-second (nominally 30 frame without outlier rejection) R,G,B image sets.

7. Combine images with 1 second cadence.  Within each 1-second bin take SNR weighted median combination of (nominally 30) R, G, and B images to produce a single image in each color (note: this process will shift color balance in later image recombination since SNR weights are different in each color).  N. B.: Special processing (sensitivity compensation) was required for the 1-second bin where the camera sensitivity (aperture) was adjusted while recording).

8. Create 3-color plane TIFF files from 1-second cadence (R,G,B) combined images.

9. Linearly re-stretch images (applied equally to all 3 color planes) to near full scale, 3-sigma, in compensation for the less than (8 bit per color) dynamic range used (e.g., dynamic range was actually very low when eclipse first became visible).

10. Adjust inter-frame color balance for color-neutral dispersion and global distribution (e.g., coronal color shifts toward  green with increasing cloud).

11. Assemble frames into a QT Movie with 1 fps playback and attach audio track.


1 - We would like to publicly thank Rick Brown for all of the hard and diligent work he had done over a very long time on all of the logistical and planning details that made this "eclipse chase" a success despite the intrusion of somewhat uncooperative weather.  Thanks, Rick, for an unforgettable 94 seconds.

2 - * 22 June 2009 will forever be a black day in the mind of this eclipse-chaser, and not due to the passing of the Moon's shadow (which would not happen for another  month), but due to the passing of my favorite photographic emulsion as proclaimed by Kodak. Quoting Paul Simon -- in TOTAL agreement:
 They give us those nice bright colors
 They give us the greens of summers
 Makes you think all the world's a sunny day, Oh yeah
 I got a Nikon camera
 I love to take a photograph
 So mama don't take my Kodachrome away"

But they did. :-(

Return to Glenn Schneider's Home Page