Karen O. |
Although it might not seem that way, we did make a huge amount of progress yesterday.
Riccardo has asked to say something. |
Riccardo |
I'd like to throw out a summary of where we are and how we might proceed.
Editorial note: A copy of Riccardo's presentation is available in:
HTML,
PDF or
PPT format.
We all agree there are a number of very exciting EALFA surveys:
- ALFALFA
- Deep surveys
- Virgo
- Can Ven
- Around big galaxies
- Really deep field (1000s sec/beam)
- Deep declination strip
- Beyond ZOA
- ZOA
We agree on various ones, but they have different needs and may
not be of equal interest to everyone.
We ask NAIC to provide 200 MHz bandwidth as soon as possible.
Our governing principles should be:
- Do the best science!
- Train the next generation.
- Produce a set of legacy surveys.
- Have the broadest astronomical impact.
There are certain realities with have to deal with:
- Not all hardware(e.g. backends) may be available immediately.
- The surveys will typically take several years to complete and the
competition among them (within E-ALFA as well as across the
disciplines) will be stiff.
- An enormous amount of work is needed before the surveys actually start.
While we have been talking very freely without any consideration of
constraints, we may face limitations in terms of:
- Realistic telescope allocations
- Effort involved in survey design and execution
- Effort involved in software development
- Effort involved in data product delivery
- Competition for NAIC resources
- Support for individual participants
There are many NAIC/ALFA challenges that we need to address:
- Technical challenges:
- Survey strategies: drift vs. drive vs. stare; multipass
- ALFA scan simulator
- sidelobe/coma simulator
- Tsys/G ~3; lots of detections! (human detection not good enough)
- Beam shape/Tsys/gain variations
- Zenith angle variations
- Sidelobe/coma contamination
- RFI environment
- Extended source vs. point source
- Political challenges:
- Highly competitive time allocation
- US national center mission to serve a broader astronomical community
- N! effect (involvement of N people)
- Who is going to do the work?
- Because of its broader responsibilities,
NAIC must protect itself from delays arising within the observing teams.
Data taking/early processing
- RFI mitigation
- Bandpass calibration/subtraction
- Zenith angle variations
- Beam shape/Tsys/gain variations
- Continuum subtraction/CLEAN
- RFI excision techniques
- Astrometric calibration/pointing
- Gridding
- Raw data archiving
Detection algorithms
- Point source vs. extended source
- Sidelobe "cleaning"/removal
- Baseline fluctuations/removal
- Template fitting (Davies et al. 2001)
- Noise characteristics and their variation
- Injection of "fake" sources (Rosenberg & Schneider)
- Source catalog generation and cross correlation with other catalogs
Data Products and Access Tools
- Low level
- What, when and how?
- Bandpass calibrated/subtracted scans
- RFI excised spectra
- Firstpass maps
- Continuum source catalog
- High level
- What, when and how?
- CLEANed data cubes
- Calibrated spectra at any position
- Catalogs of detected objects
- Access tools
- Need to provide NAIC with input
- Must be decided what are commonality of standards so that different
surveys produce similar products
Product specifications
- Product definition: who, what, when, how?
- Format specification
- Quality assurance/verification process
- User interface develoment
- Documentation
- Portal interface: NVO, IVO
- Software contribution to NAIC
US University-based participants
- Catch-22: NSF PI grant playing field
- Graduation student support (tuition, location, duration)
- Commitment/enforcement of deliverables
- - AASC recommendations re: NSF funds
- - FIRST, SIRTF legacy, HST Treasury
- GBT-like grad student program, tied to survey D&D not just observing
- GBT-like instrumentation program, extended to include software
Modus Operandi Hypothesis:      
Unlike P-ALFA, E-ALFA is characterized by:
- Diverse set of science aims/interests
- In general, less hands-on AO experience
- Less processing-intensive low level data products
- More varied multi-wavelength/correlative studies
- More diverse data products
A Modus Operandi Proposal:
- E-ALFA subcommittees will articulate a set of recommendations
to NAIC on behalf of broader community, e.g. a white paper.
- A Steering Committee will coordinate the effort and recommend ground
rules for participation/membership.
- Participation in committee work will require commitment to a significant
level of engagement. Those who choose not to commit to committee work
at this point will be kept informed of proceedings in a timely
manner.
- Between now and July 1, the E-ALFA subcommittees will work on a
preliminary version of a white paper which will outline the general
survey specifications,design studies, manpower and other resources needed,
software and data needs, and standards (we all want the surveys to function
according to common standards if possible, access tools to be compatible,
especially if different surveys might be carried out by different teams).
- NAIC will provide E-ALFA with a schedule for delivery of the final version
of this white paper and will provide detailed comments on the Jul 1 draft
- By the end-of-year, there will be a call for proposals by NAIC. At that time,
the proposing teams will form (based on a better open and public
understanding of what is involved), and proposals will be submitted.
Note: there may be several survey teams.
To accomplish this, I propose that we establish subcomittees to focus on the
required tasks.
Subcommittee Proposal
1.   Survey Science, Strategies, Simulations |
2.   Phase 1 Data: rfi mitigation & excision, data calibration, and gain
variation issues |
3.  Detection algorithms, sidelobe contamination, cleaning, source extraction |
4.   Data Products, definition, access, standards, and archiving |
5.   Follow-up science, telescope(s) requirements
        (If we need another N hours of AO telescope time to
do follow-up, we need to specify early. Also coordinate for other telescopes.) |
6.   Synergies, piggybacking options |
|
Karen O. |
We have a lot of survey ideas. An alternative proposal is to take all the
different survey proposals and send them back to individual people to write up
the science cases for each. |
Brent |
A subcommitee on survey science might
be a higher level committee with subcommittees |
Liese |
What we need is a subcommittee to provide the specifications/requirements for E-ALFA to NAIC. |
Bob |
The science discussion was very interesting, but lacking in it was specificity
as to the technical requirements. We need to be very open minded; this is just what
I need. Ask: what do the science requirement mean in terms of technical requirements?
Not just hardware, but also software? Do you have specific models for how calibration
gets done? What about for baseline removal? Maybe not everyone needs to worry equally
about RFI mitigation (too technical), but maybe some cannot function without it.
Perhaps ALFALFA doesn't need to survey below 1350. But NAIC needs to understand what the
specifications are so that we can assess their impact on our long term hardware and software
development plans. |
Karen O. |
One thing that goes along with what Bob is saying is that each survey group needs to
come back with specific requirements. For example, we should ask for what we each want
in terms of bandwidth, resolution, sample (e.g. 9 level), etc. Let's review the
surveys we identified yesterday:
1. | Virgo |
2. | Can Ven |
3. | Around big galaxies |
4. | Really deep field |
5. | Deep declination strip |
6. | ZOA |
|
Noah |
I have been advocating Virgo, but let me put on my other hat. I am on the
IAU working group on surveys. We need to look at the ALFA surveys in combination
with other surveys going on in astronomy. Also, while you might split people up now
into separate groups, there will be overlap in the interests. So how do we decide
which one to participate in. Also, I am not sure that all of us are convinced
that you can get good baselines in drive mode rather than drift, so we need to
look carefully at the technical issues. |
Karen O. |
Certainly, the shallow survey is very interesting.
Maybe details of observing modes can come later, but right now, we do need to
break up into subgroups so we can set some numbers on the survey modes. |
Martha |
We probably can address the science because it is fun and relatively easy.
In fact, we did a lot yesterday. But other tasks are more complicated
and a whole lot less fun, but they also need to be done to make progress with ALFA.
And they need to be done soon. We need some organization to figure
out how to develop and produce this white paper. The P-ALFA group did that before
their workshop. Perhaps it is easier for them, but we also need to do that.
They have advantage because the vast majority of the P-ALFA folks use Arecibo
a lot, and we have perhaps broader survey needs and goals. But we need to keep
in mind that we have to compete with P-ALFA for telescope time, so we have to
be sure we are in a position to do that. Whatever we want, proposals won't get
allocated time in competition if all of the details are not thoroughly worked
our. Personally, I'd like to talk about science all day, but I'm afraid we also
need to spend some time on the other gory details. |
Steve |
I propose that we spend some time having groups look at the details, make science
somewhat focused, and define the required specifications within that framework. |
Liese |
Some of the surveys could be lumped together, how about HI and environment? |
Karen O. |
We seem to have 5 different survey categories. Let's assign someone to each
category to lead the discussion.
Individuals were proposed to lead the discussion of
survey needs in 5 categories:
Survey | Leader |
HI in different environments | Liese |
Ultra deep survey | Lister |
Deep strip survey | Steve |
ALFALFA | Riccardo |
ZOA | Trish |
I propose that we break into subgroups and come back in an hour to see where
we are.
|
Bob |
Please keep in mind that we need to know what you folks require in terms of
instrumentation and software, and what will be produced in terms
of data products.
|
Lister |
Actually, this turns out to be a lot harder than we might have expected but it
would be unique.
The most compelling science is tied to the evolution of the gas density in the universe
(e.g. Pei & Fal 1999). The co-moving gas density goes as (1 + z)3.
ALFA can probe only to z to 0.1, but there seems to be very steep evolution with z,
so even over that relatively low range, we should see the effect.
Models based on multi wavelength observations predict that the density
goes as (1 + z)3.2.
Observations of damped LyA absorption lines suggest that the number of
absorbers per unit redshift interval goes as
        dN(DLA)/dz = 0.05(1+z)1.1,
implying essentially no evolution for DLA absorbers (Storrie-Lombardi & Wolfe 2000).
The goal of the Ultra Deep Survey would be to differentiate between evolving
and non-evolving HI density. To discriminate between the two models, we would
need to be able to discern a 50% change in density at z=0.15. The survey would be designed
to detect 40 galaxies with masses of 109.5 solar masses
at the 99% confidence level. That requires about 0.05 mJy/beam, about 260 times deeper
than HIPASS. This would be unique science that only Arecibo can do.
We could cover 0.36 square deg in about 1000 hours, say ~70 hours per beam;
this would correspond to a volume of 8000 Mpc3 and would give
160 galaxies in the range 109- 1010 solar masses. A by-product
would be a survey of redshifted OH megamasers.
This survey would probably be done in drift scan mode, amounting to
1000 hours at 3 hours/day, so it is certainly a long term project.
- Required technical parameters for an Ultra Deep Survey
- 200 MHz bandwidth
- Multilevel sampling (9 minimum, preferably more, like 14-bit)
- 8192 channels
- Fast dumping to excise RFI
- Radar blanking, or very fast dumping (~1 ms)
- Nighttime observations (though willing to test daytime)
- Need calibration and beam parmameters for all zenith angles
- Best would be declination near 17.5 deg or 19.5 deg and |b|>30 deg.
- Could be near some existing deep field
- Software
- Must be capable of dealing with drift scans
- Data Products
- Data cubes
- Source catalog
- Deep optical followup
- Overlap with other surveys
- Need HIMF from shallow surveys, for comparison with low z
- Good test of ultimate performance of ALFA.
|
Ed |
In our discussion of a survey of Virgo, we were talking about observing
a few regions for long integration times also.
Could you do these two piggybacking? |
Liese |
Having a foreground galaxy might affect the ability to do optical follow-up. |
Riccardo |
If off the edge of galaxy, then? |
Ed |
I would be very interested in those observations then. |
Jon |
Could you include a quasar so you could also look at absorption? |
Liese |
Can you make it an optical quasar, not a radio one? You don't want continuum. |
Karen O. |
Yes, we don't want baseline problems from any radio sources. |
Steve |
First, I have to say that I would like to have more time to understand better the
coverage strategies and the simulations such as Riccardo presented.
We need to consider the differences in the different beams (the center beam versus outer ones)
in setting the survey strategies. I have questions about how much integration time
is needed. I think they might be up to a factor 2 longer.
The goal of a deep strip is to go order of magnitude deeper than any previous survey.
Therefore we would like 10 sigma detections of 106 solar masses
out to 10 Mpc across a range of environments. A summary of our needs is:
- Proposed is a narrow strip, probably out of galactic plane, 1 degree x 300 degrees.
- Mode would be drift scans, stepping by a single beam so a single pixel crosses
each sky pixel to get uniform scan.
- 80 sec per beam total.
- We'd want a pretty simple data product: data cubes + a catalog but not a unique one.
then different people could apply different detection algorithms to it.
- Basic FITS format; that is pretty straight forward.
- We would prefer not to do radar blanking. We might cut down to 50 MHz to reduce the
RFI. That would be more efficient (overhead of blanking) and safer.
We'd want to allow for repeated coverage for RFI identification.
- We want 9 level sampling, and may want 2.5 km/s so we can hanning smooth.
- We felt there is o pressure on this survey to go to very high z.
- Roughly need 600-700 hours to cover this area.
|
Chris |
Radar blanking only adds 15-20%. |
DJ |
Does daytime vs nighttime matter? |
Steve |
There has been some sucess with drift scans during the daytime,
doing fourier subtraction of standing waves. So daytime may be possible. |
Noah |
Any specific declination you want? |
Steve |
We probably want to avoid Virgo to not cover it again. Likewise, we'd want
to avoid the galactic plane. But, no particular spot is favored right now. |
Trish |
In discussing this, we tried to think about what we would we do if we were NOT
piggybacking. ZOA science is not to map terribly distant structures, but rather
to look at relatively nearby structures, especially for those which are responsible
for the motion of the Milky Way.
So the basic requirements are pretty simple:   100 MHz, 3 level is ok, 5 km/s resolution. |
Liese |
But isn't this exactly where you'd want to do something that optical surveys
cannot do? |
Trish |
That would be nice, but not really essential for mapping the masses that are
pulling around Milky Way. |
Lister |
The high z universe is statistically the same everywhere, so it's not so important
where you do it. |
Liese |
But you want to add something that optical surveys cannot touch? |
Trish |
Only if you think its critical to map structures behind the ZoA at large z. |
Karen O. |
Don't you care about 9 level sampling? With anything less, you get more ringing '
even in 100 MHz. I would recommend 9 level sampling for interference mitigation. |
Trish |
This survey not need terribly deep. A few 100 seconds per beam would be great.
What is critical is positional accuracy because of the lack of optical information.
So, the mode would be drift scan, Nyquist sampled. 1 pass would require 330 hours but
a second pass would require of order 700 hours. A double pass of an ALFALFA style survey
but with longer integration time would be good.
The goal is to get a factor of 2 below M* at 100 Mpc. |
Liese |
How much deeper would this be than HIPASS? |
Trish |
5 mJy in one pass and 25 kHz resolution, so you could smooth. |
Karen O. |
Much better velocity resolution? |
Lister |
The Parkes Deep ZOA survey goes to about +20 deg and gets 5 mJy. |
Jess |
Riccardo has already discussed in some detail the ideas and motivations behind
a shallow all-sky survey.
- 200 MHz would provide more science, but not primary science
- How to go about doing this, drift vs drive
- Drift scans clearly way to go, but a lot more time required to do it.
- Would take more time, so perhaps driven scans should be done.
- Ideal survey would do whole sky, 3 passes in drift.
- Data product: fairly simple
- also need a record of what is done to data
- RFI: A multipass mode would not only provide transient detection (of interest
to others) but also help with rfi identification.
|
Mary |
Did you think about combining with a HVC survey to use the 200 MHz to
get higher velocity resolution? |
Jess |
That is another option to get higher resolution. It certainly needs to be discussed.
Our primary criterion would be 5 km/s but higher might be useful. |
Karen O. |
A lot of discussion has gone one about the importance of getting the low mass end
of the HI mass function. How many more objects can you get with ALFALFA than HIPASS did?
If you don't make a big enough improvement,
than it might not be worth the investment of telescope time. |
Jess |
Riccardo did some of that in his presentation. Also, ALFALFA would have a lot
better spatial and velocity resolution than HIPASS. |
Riccardo |
A question for Jim? Would you rather have single drift scan (12 sec) or
3 passes? |
Jim |
Personally I would vote for multi passes (for transients), but
this is clearly not as high a priority. |
DJ |
How much time would be required for this survey? |
Jessica |
The original idea was 1200 hours, but it depends on the final adopted strategy. |
DJ |
Does daytime vs nighttime matter? |
Riccardo |
We would prefer nighttime but part of day could be used.
We would want to avoid noon, sunrise and sunset, because of baselines. |
Bob |
Let me lay out what I see it is that NAIC needs from you.
If we think about who the stakeholders are in this project, NAIC is one of the
stakeholders. This project will be a legacy of the Observatory, so it is
clearly critical to us that ALFA and its surveys be successful.
But, you are all stakeholders in this too. You see the goals, so we view you as
critical to the success. Also, we must include among the stakeholders
the people who are not here. It is very important to include all of the stakeholders
in the discussions.
So, whatever initial plan you come up with should be discussed with the
wider community. We need to insure that the final product is something
that everyone will be happy with.
There are 3 stakeholders in this enterprise: (1) NAIC, (2) those of us here today,
and (3) the others who aren't here, but are nonetheless interested. All three groups
need to be happy with the outcome. Some we must make sure there is plenty of
opportunity for open discussion.
So, the big question is: What are the next steps? Let me propose some:
- try to understand synergies
- muster the resources
To go forward, we need to catalog the discussion, not just today, but as it
develops. For that, we need a white paper (or whatever you want to call it).
So, we need to lay groundwork for the whitepaper, both to get information out
so it can be widely discussed and to advise NAIC.
So, what might the white paper look like? It's not just
to define the science surveys but also to define where we are at the moment
and to present ideas to others who might not be here now. Let me propose
as possible outline, which we can then discuss.
Outline of the E-ALFA "white paper" - status/plans report
- Science justification for surveys
- Observing strategies
- need to be fairly specific
- might involve survey simulations
- Early data products (NAIC responsibility)
- Raw data clearly kept
- Phase I: basically NAIC is responsible to get data off the
telescope and into archive
- Calibrated, passband corrected, RFI excision (?)
(more than the raw data, feeds next step)
(white paper needs to outline what/how)
- Source extraction algorithms
- How will it be done, what options, what methods
- Phase II data products and standards (Community responsibility)
- Deconvolution, clean data cube, source catalog
- Archive and Access
- Follow up observations
- Do we need VLA, optical or more AO follow-up? What do we expect to need?
- Synergies with other consortia
We need this white paper in timely manner so that NAIC, the other consortia
and the rest of the community can begin to understand what's coming with ALFA so
that we are all able to plan for it.
Since the white paper needs to get written, let's see how we can organize ourselves
to get that task done. Following the outline, then let me propose a possible organization layout.
(Click here for a bigger version of the org chart.)
- Science and strategies (mostly community)
- Science justification
- Survey strategies
- Phase I data/products (primary NAIC involvement)
- Source detection and deconvolution (community responsibility)
(A lot of work, a lot of thought to achieve science)
- Detection algorithms
- Phase II data products
- Follow up science (might come under science as well)
- Synergies with other surveys
|
|
The organization then:
- Needs leaders to make sure that each box gets written
- Needs leader(s) for steering group
... so that a document gets produced that contains these items and then
gets distributed
|
Steve |
Phase 1 data reduction should be reversible, so I disagree with having automatic RFI removal. |
Bob |
The raw data won't go away. NAIC will archive it. |
Liese |
Does Phase 1 data reduction include deconvolution? |
Bob |
No, that belongs in the community domain. |
Liese |
Will NAIC help us characterize the beam pattern so that we can undertake the deconvolution? |
Bob |
Yes, of course.
|
Karen O. |
Your draft white paper does not have anything about membership criteria, what is a
consortium etc. |
Bob |
Yes. You need to decide how those issues are going to get written. That is one of the basic
points. |
Riccardo |
Isn't it necessary for NAIC to be involved in the definition of needs and requirements
and what NAIC expects? |
Bob |
Very much so. NAIC needs to understand how the surveys will function, but also the
community needs to understand what they are committing themselves to and what
NAIC expects of the teams. |
Riccardo |
Also, there may be different surveys, so a set of standards needs to be set so that
final data products look the same in the end. It would be more useful to have an effort
that might bridge across surveys/disciplines, so that the final products
have a common face to the public. |
Bob |
We want to get a draft sooner than later but it should be an iterative process. |
Martha |
We want to figure out how to allow communication of technical expertise amongst us. |
Karen O. |
Only half of the people who signed up for E-ALFA are actually here. |
Bob |
How do we establish community outreach, including community support, especially for those
supported by NSF? |
Riccardo |
What is the timescale for a draft? What is NAIC's involvement in putting this together? |
Bob |
The sooner the better. I don't know the precise answer because I don't know how
difficult it will prove to be. |
Wim |
Would 3 months be reasonable? |
Bob |
That seems reasonable. |