Friday, January 28, 2022

Ayios Eustratios on Crete (C7416)


In the years 1371 - 1374 the Venetians built a castle, Frangokastello, on the S coast of Crete.[1]  I can't resist giving you a picture of it:

The general area in S. Crete looks like this:

Near this castle there was, in Roman times, a (port?) town called Ayios Eustratios.  The ruins of that  town can still be seen although these days it is nothing but a series of collapsed walls, completely abandoned except for someone who seems to have hacked out a farm there. 

The question comes down to where is Ayios Eustratios exactly; Pleiades' position is catastrophically wrong as usual.  How do we find it?

The best description of Ayios Eustratios that I can find is in Hood who says: 

"Extensive Roman settlement in the middle of the plain about 1,500 metres north-west of Frangokastelli and less than a kilometre from the sea."[1]

He provides a sketch map although it is useless.  Here it is:

First let's look for the little chapel of Ayia Pelayia.  It is located at  35.188103° N  24.224004° S (F6754) and here's a photo of it from Google Street View:

The difficulty with Hood's map is that it places the ruins of Ayios Eustratios on the same side of the coastal road as Ayia Pelayia and, from Hood's verbal description, this is not possible.  In addition the coastal road is shown on the map much further inland than it is in reality.

Hood wants these ruins to be 1500 m. from Fragokastello and 'less than' one km. from the coast.  Perhaps we can bound the area of uncertainty by using these constraints.  The next picture shows a white bounding box which covers the area 1500 m. from Fragokastello Castle  (red circle) and a kilometer from the coastline (the yellow line):

Somewhere in here should be the ruins of Ayios Eustratios.  But where?   I wanted more clues so I headed back to Hood who gives a photograph of the apse of a ruined church in the Ayios Eustratios.  It looks like this:

This is a column from the apse of the church of Ayios Eustratios.  The only clue which this picture affords is a snippet of ridge-line behind the church on the left.  If it were possible to locate this ridge line it would be possible, just, to draw a back azimuth that might cross our town.  Is that possible?  First I wanted more detail in the ridge-line.

Well, this looks  pretty bad.  It's not going into our collection of vacation snapshots.  But that's not the point.  I ran this through an HDR processor from NIK software and crunched it to the maximum.  Now it's quite possible to make out more detail in our ridge-line.  However, we still haven't found the snippet of ridge that this shows.

I went back into Google Earth and tried to locate this ridge.  Here it is from Google's ground view:

If you compare the enhanced photo with the ridge shown in the white box you'll see the same section of ridge-line.  Just to be more certain I looked at the same ridge from Google's Street View:

I enhanced the ridge snippet from the Google Street Drive photo and got this:

On the left is Hood's picture and on the right is the enhanced section of ridge from Google's Street View.  I think that this is it.  Now if we can draw a line from that ridge point and at right angles to it, through the bounding box we might better be able to locate our ruins.  Here it is:

In this picture the purple line is drawn from the top of the ridge in the photo at what, frankly, is a guessed angle through the bounding box.  The red circle is radius 1500 m. from Fragokastello Castle.  And the yellow line is about 1 km. from the coast.  All the constraints are met so somewhere, at the NE side?, of the white bounding box we should be able to find the town of Ayios Eustratios.

Here it is in close-up; what's that semi-wooded area at the right corner?

This looks like our town.  Hood says: 

"The site is strewn with abundant stones from buildings, and the ruins of an early Christian basilica church are visible."

I ran this picture through every algorithm I could think of in order to bring out the ruined walls and here they are, unmistakably.   The former walls have the curious 'smeared' appearance of walls that have sat undisturbed for a couple of thousand years.  I am convinced that this is the right place.

Can we now find the church of Ayios Eustratios?

I think that this shows it.  The structure pointed to by the arrow is a little longer than Hood says, about 18  m. vs. his 15.  But the width is right.  I make it to be just a hair more than Hood's 12 meters.

 The real scandal here is how consistently bad Pleiades' data is.  Here's a map which compares my position to theirs.  The length of the red line that joins the two positions is 2122 m. long.  So Pleiades is more than two km. out.  Tell me again why Pleiades' data is used by scholars?

Ayios Eustratios is now in the M.A.P. as C7416.


[1] Hood [1967] 55,   'B. 7. Ayios Evstratios'.


 Hood [1967]:  Hood, M.S.F.  'Some Ancient Sites in South-West Crete',  The Annual of the British School at Athens, (62) pp. 47-56. 1967.  Online here.  

Monday, January 24, 2022

 Pleiades and Roukouni Korphi

Sometimes I wonder why Pleiades even bothers.  In truth, I wonder that a lot of the time.

Case in point: Pleiades no. 590035 which has the name 'Roukouni Korphi'.  After clicking on this and getting the inevitable annoying zoom-in from outer space we find ourselves staring at a barren landscape somewhere in SE Crete. 

What is Roukouni Korphi?  We are not told. 
Why was it selected for the atlas?  We are not told.
What does it depend on?  We are told that it is digitized from the Barrington Atlas

Well, it turns out that Roukouni Korphi is a circular structure,, perhaps a tower or look-out of some sort that dates from the Hellenistic.  I discovered this long afterwards.  The immediate problem was to find it because Pleiades' coordinates are obviously wrong.

I discovered that the ultimate source for this was in an article by Sinclair Hood who describes it like this:

"(11) ‘Roukouni Korfi’ ('Ρονκούνη Κορφή). Α high peak at the western end of the range of hills
on which (6) ‘Xinakhladha’ and ‘Tourkissa’, (7) ‘Kastellos’, and the sites (8), (9) near it lie,
about five minutes above (7) ‘Kastellos’ to the west. A few metres east of the summit Platon
has excavated an interesting circular building, which appears to be a beacon or guard station
constructed in Late Hellenistic times (Ergon for 1960, 206 f., with plan; cf. BCH lxxxv (1961)
876—7; ΚΚ xiv (1960) 511—12; ADelt xvi ( 1960) 259). This was no doubt in the territory of
ancient BIANNOS (see p. 83), and from the summit there is a very extensive View over the
Mesara plain to the west with Tsoutsouros (ancient INATOS) to the south, and as far as Arkalo—
khori and Mt. Juktas to the north."

So all we have to do is figure out where Xinakhladha, Tourkissa, Kastellos, and sites 8 (Amigdhaloi) and 9 (Bubouli) are located and then we look for a 'high peak' near there. I guess.

I went over Hood's article iwth a fine-toothed comb and I plotted all the points which he and his team describe around Khondros. The result is as you see in the next picture:

This does look like a section of a Jackson Pollock.  In actuality it is the area just above Kastri, NE of Inatos and SW of Biannos in south-eastern Crete.  There are two towns here, Perivolia on the upper right and Chondros on the lower right.  To the W of Chondros there is a valley which is watered by the Chondros stream.  On the S of that stream is a range of rocky hills on the top of which lies Roukouni Korphi (arrow 'Actual' at lower left center).

Here's another view from the north that shows the area in question:


Hood and his team name 11 sites worthy of note in this area of which R.K. is no. 11.  In order to locate it securely I located all 11 as closely as the descriptions permit.  After having done that it was relatively easy to locate R.K.  It is described as " ... about five minutes above (7) 'Kastellos' to the west."  The lat/lon pair for Kastellos is: 35.026222° N, 25.364229° E   When I scanned the peaks just to the W of 'Kastellos' I quickly discovered this: 

This matches the diagram given in Daux [1961].  [2]

Its position is: 35.023444° N, 25.359333° E

So, relative to its true location, where did Pleiades put it?  I measured the distance between the actual location and the Pleiades given location.  It turned out to be 949 m, which is almost exactly the median predicted Pleiades error of 930 m.[3]


[1] Hood, et al. [1964].
[2] Daux [1961] 873.  The description of the site is given on pp. 876-7.
[3] Consoli [2013].


Consoli [2013] : Consoli, Robert,  'A frequently asked question about the Pleiades redesign', Squinches (a Wordpress Blog),  7/24/2013.  Online here.

Daux [1961] : Georges Daux, 'Chronique des fouilles et découvertes archéologiques en Grèce en 1960', Bulletin de Correspondance Hellénique [85], pp. 601 - 953.  1961.  Online here.

Hood et al. [1964] : Hood, Sinclair and Peter Warren , Gerald Cadogan, 'Travels in Crete, 1962', The Annual of the British School at Athens (59), pp. 50-99. 1964.  Online here.

Tuesday, January 18, 2022

Ceramic Horizon Search and Years AD


New Ceramic Horizons search page

The Mycenaean Atlas has a new way to search for date ranges if that's all you want to do.  Right now you can do this from the Control Page drop-down menu for 'Ceramic Horizon'.  I thought that it would be handy to be able to search for time periods by thumbnail, the same way that you can search by region from the thumbnail maps.  So I created such a whole page search just for ceramic horizons.  To the Control Page thumbnail box I added a new thumbnail called 'CHRON'. 

When you click on that thumbnail you'll get the new Chrono search page:

It should be easy to understand this page.  The abbreviations for the various ceramic horizons are standard English abbreviations.  When you click on one the Atlas will return a map page with just the sites on it for that ceramic horizon.  At this point you can generate reports for that ceramic horizon if that's what you want to do.

The page does not cover all date ranges (ceramic horizons).  For the full set you have to go to the drop-down.  There ahould be enough choices on the new ceramic horizon page to make it very useful.

This page does not yet include a number of useful ceramic horizons, specifically 'Cypriot' but I'm going to work on adding those.

Years AD

I've been experimenting with the idea of bringing coverage in the Atlas forward in time from the Bronze Age.  The island of Naxos is a convenient place to start as it is self-contained.  The first of these sites to be added was Irokastro (C7405) which was a stronghold  in the SE part of the island.  

The page for C7405 looks like this:

Adding this site caused breakage to the Atlas because C7405's dates of use span the year 1 AD and the database isn't designed for years in the Christian era.  What to do?  I've defined the R(oman) period as 753 BC to 395 AD. 

  The old time_span DB table has two columns for year ranges: post and ante.  If I just stored R(oman) as post:753 and ante:395 the software would interpret this as R ranging from 753 BC to 395 BC.  Not good.

   This is a fairly vanilla problem in DB design and development and several solutions are possible.  First, as doctors say, do no harm.  The least desirable solution is to perturb what's already there.  In its restricted range of BC the code already works fine.  How will we add in years AD in a non-destructive way?

   One thing that could be done is to assign years BC a minus sign: Make them negative with years AD being positive.  This kind of solution makes me itch and sounds too clever by half.  I don't do any year span calculation in the code but if I did this would be the wrong solution.  

Computing a range of years with the years BC negative would be like this:

$R = $Yante + abs($Ypost) - 1;

So 753 BC to 395 AD is 395 - -753 - 1 = 1147;

Making the years AD negative is still icky and amounts to about the same thing.  Here computing a range of years is just Ypost - Yante - 1.   

So 753 BC to 395 AD is just 753 - -395 - 1 = 1147 (we subtract 1 because there's no year 0).  

I still don't like it.  

   A better solution is to bite the bullet and add a column or two(?) to the DB time_span table which will carry the era information about the era for each year range endpoint.  In this way we separate out the absolute year value from the era it belongs to.  Having done that we can now rely on the software to put these together in whatever way seems suitable at coding time.  What should these extra columns contain?  Well we could have two columns of binary values; one column for post and one for ante.  false (0) could stand for BC and true for AD.  This is the classic computer solution and a lot of people would do this but from my perspective this has a drawback.  

For the two year range endpoints there are only three legitimate values: BC & BC, BC & AD, and AD & AD.   AD & BC (AD followed by BC) is not a valid value.  

Decoupling these two values into two separate columns will not automatically prevent our specifying AD-BC or 1,0.  I want to be protected from my own stupidity.   I finally decided to implement 1 new field with a set of three allowable values: '00', '01', and '11' where 0 represents 'BC' and 1 represents 'AD'.  The DB allows no other values in this field (which I called 'bcad') such as '10'.   I defaulted this field for the whole table to '00'.  I then set the new R record to '01'.

  There remained two places in the SW where changes had to be made: 
a) where I print out the year ranges in the place key report page and 
b) where I pop-up the year ranges on the Chron report page.  

I had to change the query to retrieve the new bcad field for each range and write separate code for each of the new three cases to add on the correct 'BC' or 'AD' suffix before printing out the year range:  In PHP pseufo-code it looks like this:

switch ($bcad)
Case '00'  : {$postname = $postname."BC"; $antename = $antename. " BC";  break;}
case '01' : {$postname = $postname. "BC"; $antename = $antename." AD" ; break;}
case '11' : {$postname = $postname. "AD";  $antename = $antename." AD" ; break;}
default : {break;}

or ( more cryptic)

$postname = $postname. " BC";   $antename = $antename. " BC";

switch ($bcad)
case '01' : { $antename = $antename." AD" ; break;}
case '11' : {$postname = $postname. " AD"; $antename = $antename. " AD"; break;}
default : {break;}

The first is what I actually did.  Another solution consists of treating the $bcad values as indexes into an array of suffixes:

$baArray = array( "BC", "AD");

The code is then just:

$postname = $postname." ".$baArray[$bcad[0]];
$antename = $antename." ".$baArray[$bcad[1]];

 O.k. you get the idea.  So on the place key report page you now will see 'BC' for years BC where you never did before and when you look at C7405 you'll see "753 BC - 395 AD" shown correctly:

 This also goes for the tooltips on the chronology page.  I hope that's the last correction I have to make to support dates 'AD'.