

In [1]:

    
%%html
<script type="text/javascript">
     show=true;
     function toggle(){
         if (show){$('div.input').hide();}else{$('div.input').show();}
            show = !show}
 </script>
 <h2><a href="javascript:toggle()" target="_self">Click to toggle code input</a></h2>









    





 Click to toggle code input

Mario Juric at SETI Institute January 14, 2014

LSST, Entering the Era of Petascale Optical Astronomy

Link to the full YouTube video

TL;DW:

What is LSST going to do and what are some of its challenges.



In [2]:

    
from IPython.display import YouTubeVideo
from datetime import timedelta

Overview of LSST-

The ultimate deliverable of LSST is the fully reduced data.
Robotic telescope
Publicly funded project with early contributions from private donors
NSF and DOE mostly-- so totally public, no proprietary period

"[LSST] is practically a robot that sits on a mountain and does its thing."



In [3]:

    
YouTubeVideo('n4OysNeyXpc', start=225, end=243)









    Out[3]:

LSST science goals

Time domain science- repeated observations of the sky (~800 repeated visits)
- Supernovae
- Compact object mergers
Solar system science
- Census of near earth asteroids
- Potentially hazardous
Mapping the Milky Way
- Galactic structure
- Tidal streams
Dark Energy and Dark Matter (10:18)
- Major driver of LSST
- Dark matter is cosmological constant or something else?



In [4]:

    
start = int(timedelta(hours=0, minutes=7, seconds=18).total_seconds())
end = int(timedelta(hours=0, minutes=12, seconds=2).total_seconds())
YouTubeVideo('n4OysNeyXpc', start=438, end=722)









    Out[4]:

Decadal survey ranked LSST as the top ground based facility for the next decade.



In [5]:

    
start = int(timedelta(hours=0, minutes=12, seconds=2).total_seconds())
end = int(timedelta(hours=0, minutes=12, seconds=44).total_seconds())
YouTubeVideo('n4OysNeyXpc', start=start, end=end)









    Out[5]:

Who/where is LSST?

Between about 50 and 100 people
Science case is effectively being built pro-bono by community
Location is Cerro Pachon in Chile, not far from La Serena

LSST by numbers

10 years
20,000 sq degrees
~800 visits per source
10 billion objects
Limiting magnitude of $R = 27$

LSST will not reobserve the Galactic plane.



In [6]:

    
start = int(timedelta(hours=0, minutes=17, seconds=5).total_seconds())
end = int(timedelta(hours=0, minutes=18, seconds=0).total_seconds())
YouTubeVideo('n4OysNeyXpc', start=start, end=end)









    Out[6]:

Camera will be the largest astronomical camera in existence.

3.2 Gigapixel
0.2 arcseconds per pixel
9.6 square degree FOV
6 filters
2 second readout time (!)
15 second exposure time
189 CCD chips
4k x 4k pixels
3 x 3 CCD raft (21 of these)
16 individual readouts
3.5 deg FOV
corner area wavefront sensing and guiding
10 $\mu$m pixels



In [7]:

    
start = int(timedelta(hours=0, minutes=19, seconds=37).total_seconds())
end = int(timedelta(hours=0, minutes=23, seconds=30).total_seconds())
YouTubeVideo('n4OysNeyXpc', start=start, end=end)









    Out[7]:

Data management system

Catalog generation
How do you make a petascale database to both astronomers and the public?
All the data processing is going on at the NCSA in Illinois
Data links for real-time data processing. Redundant links.
Supercomputers for data crunching.



In [8]:

    
start = int(timedelta(hours=0, minutes=23, seconds=55).total_seconds())
end = int(timedelta(hours=0, minutes=27, seconds=21).total_seconds())
YouTubeVideo('n4OysNeyXpc', start=start, end=end)









    Out[8]:

Catalogs

Level 1: time sensitive
- Time domain alerts within 60 seconds.
- What to do with ~10 million time-domain events per night?
- ~6 million bodies in the Solar System. (about 85% of asteroids bigger than 130 meters)
Level 2: stable catalog
- Deep coadds of images
- Annually released catalog
- ~37 billion objects (20B galaxies, 17B stars)
Level 3: Access to the data
- Ship users hard drives?
- Bring the questions to the data (query the data)
- Do the computing on the cloud



In [9]:

    
start = int(timedelta(hours=0, minutes=27, seconds=21).total_seconds())
end = int(timedelta(hours=0, minutes=36, seconds=54).total_seconds())
YouTubeVideo('n4OysNeyXpc', start=start, end=end)









    Out[9]:

Software

Freely available and open source (now!)
Works for Mac OSX 10.7+
Still in development

Astronomers are not used to working with machines like this.



In [10]:

    
# Software
start = int(timedelta(hours=0, minutes=40, seconds=0).total_seconds())
end = int(timedelta(hours=0, minutes=41, seconds=12).total_seconds())
#YouTubeVideo('n4OysNeyXpc', start=start, end=end)

# Effectively using large survey data
start = int(timedelta(hours=0, minutes=40, seconds=0).total_seconds())
end = int(timedelta(hours=0, minutes=46, seconds=30).total_seconds())
#YouTubeVideo('n4OysNeyXpc', start=start, end=end)

# Serendipitous science
start = int(timedelta(hours=0, minutes=46, seconds=30).total_seconds())
end = int(timedelta(hours=0, minutes=47, seconds=14).total_seconds())
#YouTubeVideo('n4OysNeyXpc', start=start, end=end)

Thriving in the Era of Large Surveys

Data driven astronomers
What theories can I test with the data I have?
Teaching the computer to find discoveries for us
Code, statistics, interfacing with large databases
Machine learning, artificial intelligence



In [11]:

    
# Thriving in the Era of Large Surveys
start = int(timedelta(hours=0, minutes=47, seconds=14).total_seconds())
end = int(timedelta(hours=0, minutes=49, seconds=50).total_seconds())
YouTubeVideo('n4OysNeyXpc', start=start, end=end)









    Out[11]:

Software engineering will become as essential as mathematics

Data driven astronomer
Software is not taken as seriously in astronomy
Software to be taught not learned through osmosis
SQL
More computer science is needed



In [12]:

    
start = int(timedelta(hours=0, minutes=49, seconds=48).total_seconds())
end = int(timedelta(hours=0, minutes=51, seconds=50).total_seconds())
YouTubeVideo('n4OysNeyXpc', start=start, end=end)









    Out[12]:

End of talk. Question time.

Calibration

Self calibration
Reobserve the sky with different parts of the detector.
Calibrations of SDSS were not used at all.
There is more data in the survey itself



In [13]:

    
start = int(timedelta(hours=0, minutes=53, seconds=50).total_seconds())
end = int(timedelta(hours=0, minutes=54, seconds=50).total_seconds())
YouTubeVideo('n4OysNeyXpc', start=start, end=end)









    Out[13]:

What and how to teach our students?

Depends on individual institutions.
Large surveys know this.
Share code
What is needed to make LSST a success?



In [14]:

    
start = int(timedelta(hours=0, minutes=59, seconds=33).total_seconds())
end = int(timedelta(hours=1, minutes=1, seconds=37).total_seconds())
YouTubeVideo('n4OysNeyXpc', start=start, end=end)









    Out[14]: