Hi Jacek,
To the best of my memory, in the 15+ years that Kem and I have worked
together I have never calculated a number that he has agreed with, so I
am used to this :-) My comments are interspersed below.
Looking at your spreadsheet, I think one problem is that you have the
whole thing growing as a power of 0.56, because you have made the
time-dependent size a fixed fraction of the deep size. Instead, one
grows linearly and the other roughly like the square root. I haven't
yet had the time to completely understand what you've done.
Tim
Jacek Becla wrote:
Keywords: DataAccWG
Hi All,
I'm trying to do a bottom-up estimate of the catalog sizes, using
Tim's numbers and input from Kem and Sergei. Attached, please find the
latest draft. It is worrisome that the numbers we are getting are much
smaller than the bottom-up estimate of 300 TB for the first year (I'm
currently getting 15 TB).
Below I am including relevant comments from Kem. Any help with
correcting these numbers would be appreciated!
thanks,
Jacek
-------------------------------------
150,000 objects/square degree seems pretty high for a single image at a
reasonable galactic latitude. I believe this is the final number of
galaxies in the 10 year stack and I don't think the first DR will have
quite that many.
There are only about 8 unique square degrees per image for hexagonal
close packing, but there will probably be a few more than 3000
fields, maybe as many as 3100.
There is something wrong with the scaling on the number of galaxies.
The formula must not be accurate across the whole range. The
(20DR)**(1.25*.45) says that compared to a single visit,
the final stack will have 29x as many galaxies. A single visit gives
about 20 galaxies per square arcminute (if I remember correctly), which
means the final stack will have 580! per square arcminute or 42 billion
for the 20,000 square degree survey. This is a factor of 10 too high.
I defined a DR as once every six months, so we have 20 DR's over the
length of the survey. This says that the number of galaxies grows by a
factor of 5.4 from DR=1 to DR=20. I calculate the limiting magnitude
for galaxies in DR=1 to be about 26.1 from stacking about 20 R band
images. The Subaru Deep Field numbers, which I also used, give a
density of galaxies at 26.25 or brighter to be 1.8e5 per deg**2. Note
that this is already 50 per arcmin**2! Multiplying by 5.4, this says
that the density should be 9.7e5 per deg**2, or 270 per arcmin**2. I
agree that this number is too high, but not by a lot. Based on my
simple model, the limiting magnitude of the final stack should be about
24.5 + 1.25 log10(200) = 27.4. I make out the SDF cumulative number at
this mag is about 3.9e5 per deg**2, not 9.7e5, so I will plead guilty to
a bit more than a factor of two. Kem says below that he gets 8.0e5, so
that would say our disagreement is a factor of 1.2. Certainly there is
no factor of 10! So, I think we are talking about at most a factor of
two in the numbers I originally sent out, and probably less.
I think there are a few things contributing to the factor of two. The
main one is my use of a constant slope in d(logN)/dM. It isn't
constant, and actually varies between about 0.2 to 0.5 over the
magnitude range. A more careful job would just incorporate the full
table into a spreadsheet and interpolate.
Unrelated to these points, in the expression above, the 20 should be a
multiplier in front, not raised to the exponent.
There is confusion in the accounting between visits and exposures.
There are going to be two exposures per visit, so there will be twice
the number of 'observations' which are relevant to the time dependent db.
I think the fraction of variable objects per field is way too high,
particularly as the number of objects increases with increasing depth.
Most of the deeper objects will be galaxies and the fraction of agn, SN
whatever in galaxies is more like .01 than 0.1.
I agree that this fraction is too high (this one didn't come from me!),
but it shouldn't affect the size numbers! Any object that has useful
S/N in a single visit should have entries in the time dependent
database, whether it is variable or not.
----------------------------------------------------------------
I just checked the Subaru Deep field for some numbers (Kashikawa etal
astroph/0410005 table 2)
I find that the total number of galaxies per square degree to R=24 which
is r = 24.5 for a (g-r) =0 galaxy (( R = r - 0.51 -.15(g-r))) is
roughly 62836 which yields 17 per square arcminute, but their data
only start at 18.25, so this may be a (very) small underestimate.
I interpolate to about 190,000 galaxies per square degree for DR1
(R=25.6 or r= 26.1)
I find that the d(logN)/dmag over the range 24 - 27 varies from .27 -
.35 and that the cumulative galaxy counts to R=27.25 (as deep as they
go) is 804329 per square degree which is about the final depth of the
ten year survey (SRD has r=27.8).
So, my earlier email using numbers from memory was not right, but then
the table numbers are not right either.
----------------------------------------------------------------
I realize that those numbers may not be quite right depending on how the
bins in table 2 are labeled. Those numbers assume the bin label refers
to the bin maximum, but the labels probably refer to the middle of the
bin. So those numbers are a little high. using the middle of the
bin, I get:
R = 24: 47097 / square degree (r = 24.5)
DR1: 150,000 / square degree (hmm, guess the number in the table is
for DR1)
10 year survey gives about 760,000 / square degree
_______________________________________________
LSST-data mailing list
[email protected]
http://www.lsstmail.org/mailman/listinfo/lsst-data
begin:vcard
fn:Tim Axelrod
n:Axelrod;Tim
org:;Large Synoptic Survey Telescope
email;internet:[EMAIL PROTECTED]
title:Data Management Project Scientist
tel;work:520-322-8735
version:2.1
end:vcard
_______________________________________________
LSST-data mailing list
[email protected]
http://www.lsstmail.org/mailman/listinfo/lsst-data
