XPost: alt.astronomy   
      
   In my wanderings I came across a couple papers reporting odd things   
   seen in the sky via ground-based telescopes.  The interesting aspect   
   was the plates being examined dated back to the early 1950s when   
   things like Starlink or even Sputnik had not screwed up the sky. Yet   
   what was reported from a couple different groups was lines of "dots"   
   and other formations that seemed to look like highly reflected   
   earth-orbiting "satellites" seemed to be spotted from time to time   
   over the years.   
      
   Typical scenarios involved 2 plates where careful examination found   
   several bright points on the older plate yet the same positions in the   
   newer plate were empty.   
      
   "Careful" includes checking for faults in the emulsion or processing   
   that may have accidentally drawn a line of dot on the negative.   
      
   One paper reported ~100 of these "disappearing stars" and a later   
   paper reported ~700 of them.   
      
   As usual, attempts to contact the authors of any of the papers have   
   resulted in 0. Some of the papers are a few years old and since the   
   pandemic I have to presume people have gone onto new careers at   
   Twitter. Or something.   
      
   So I threw the problem of finding the data and explaining what was   
   going on to a couple of AI-based programs I run from time to time.   
      
   Gathering the data was fairly straightforward. The s/w latched on to   
   the Mt Palomar Sky Surveys. The first one started around 1950 and ran   
   for 10-20 years. There was a 2nd Survey starting in ~1980 but,   
   apparently, that was never finished. There was another attempt more   
   recently, but that seems to have left no imprint on a public server   
   the programs can find.   
      
   But the 1 and 2 surveys allow us to look at mns of stars on 16x16   
   arc-min images and compare the state of most parts of the sky from the   
   1950s with the 1980s.   
      
   I already have some tools to look for "things" in images from the old   
   Kaggle days (kaggle.com/kymhorsell1) and using techniques like "strong   
   statistical features" (aka SURF) the programs managed to isolate about   
   150 objects that could be seen in the 1950s survey but had apparently   
   disappeared by the 1980s even allowing for some proper motion that may   
   have taken a star off the edge of a plate and onto an adjoining one.   
      
   The SURF parameters allow the s/w to build up a small expert system   
   that not only uses the position of the center of the "star" on the   
   plate but various qualities relating to how the individual pixels in   
   the blob are arranged -- a kind of optical fingerprint.  The s/w can   
   determine whether there is a density gradient inside the blob, and it   
   keeps track of regions on the images that are (extra) bright as well   
   as circular regions that are extra dark.  The number of pixels also   
   gives an indication of the object's brightness and the spacial   
   distribution from the object's center pixel can give an idea of   
   whether it is an extended object and/or is interacting with a nearby   
   object that is too close for the relevant telescope to actually   
   resolve. It is also possible to determine from the pixels -- given   
   some of the exposures were upto 90 mins -- whether the object was   
   moving in a way not entirely consistent with a star. We might go into   
   that in a later post. :)   
      
   As well as "stars" that seemed to appear on earlier plates but later   
   were not present in the expected positions, the opposite was also   
   observed -- sometimes objects appeared on later plates that were not   
   there in the 1950s. Of the couple 100 "differences" noted in the   
   collection of images snaffled from the Palomar collection about 1/3   
   were of the "appearing star" type -- the rest were the regular   
   "disappearing" kind.   
      
   So with a collection of dates when an object was seen or not seen, the   
   location of the star at some point, and the various photometry   
   parameters the image processing dredged up we can run the usual massive   
   search of various data-files in my database to see whether any known   
   weather, planetary, demographic or other data on file varies on a   
   statistically similar way to any of the "disappearing star"   
   parameters and, if so, which are closer and maybe the "most likely   
   explanation" of what is going on.   
      
   And the top results off the rank are as follows:   
      
   Data-series             Lag     R2   
                           (y)   
   Mwind-GA                5        0.88518841   
   Mmaxwindseg150          5        0.88297516   
   Msduah_globe6NHLand     3        0.88092115   
   Mmaxhail-LA             4        0.86669351   
   Mwind-FL                4        0.86234448   
   Mmaxwind-GA             5        0.86232464   
   Msduah_mtSoPol          3        0.85737059   
   Mmaxuah_mtUSA49         3        0.85393912   
   Msduah_mtNH             3        0.84748392   
   Mforestpct              3        0.84629941   
   Msduah_mtSoPolOcean     3        0.83581867   
   Mminuah_mtNoExtOcean    5        0.83540929   
   Mwind-AR                4        0.82180789   
   Mneptune-zh             5        0.81655511   
   Mmaxwind-TX             3        0.81102430   
   Mmaxneptune-r           3        0.80945403   
   Mmaxhail-NJ             0        0.80611332   
   Mhail8                  4        0.80494563   
   Mhail-WY                3        0.80367691   
      
   The s/w has determined that any data series that has an R2 > .8 is   
   effectively in the "best explanation" bucket. So the above list is not   
   really an ordered list -- all items in it are "equally likely   
   explanations" the predict the coming and going of the objects in question.   
      
   We immediately see most of the explanations seem to involve the   
   weather someplace on planet dirt. The "wind" items are NOAA-reported   
   windstorms seen over the years across the US.  The "max" prefix   
   indicates the s/w took the maximum of monthly numbers to produce an   
   "annual maximum" series from the original data.  Similar for "min" and   
   "sd" (the stddev of the months or an est of annual   
   variability). Series with a numeric prefix refer to a specific   
   month. I.e. "hail8" is number of August hailstorms reported to NOAA,   
   year by year.   
      
   In the middle of the pact is one oddball -- "forestpct".  This is the   
   surveyed% of the planet that is still covered in   
   forest. Obviously this number has been going down for several decades,   
   but after WWII it was going up a little as people relaxed and started   
   planning their future families.   
      
   And, finally -- and the only thing really connected with things "out   
   there" -- the position of Neptune seems to have some predictive power   
   for when "stars disappear". The "-r" series is Neptune's distance from   
   the sun. The "-zh" is Neptune's position in AU above the ecliptic.   
      
   So apart from "probably related to earth weather" the AI programs   
   point out a possible link with Neptune. Other planets and moons also   
   turn up way down the list (there are 1000s of items in the list) but   
   Neptune is in the "best explanation" section -- the other things are not.   
      
   And finally, finally we might ask are any other (cough) odd things   
   maybe connected with "stars" popping up on telescope images some times   
   and not other times. Given some of the objects imaged *seem* to be   
      
   [continued in next message]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)