LIST Discussion Paper Planning for LISA data analysis

1
LIST Discussion PaperPlanning for LISA data
analysis

• Bernard Schutz
• with input from C Cutler, S Phinney
• December 2003

2
Summary

• Analysis of the LISA data should be recognized by
  the agencies (ESA and NASA) as an integral part
  of the project, and planning for it should be
  done as carefully as for the hardware. The
  success of the entire mission is put at risk if
  the analysis is done poorly. The analysis task
  for LISA is likely to be significantly more
  complex and demanding than that required for data
  from ground-based detectors.
• LIST should develop during the next year a
  detailed plan for data analysis, based on a
  working model that the analysis should be
  coordinated by two cooperating analysis/science
  centers, one in the USA and one in Europe. These
  centers should have the responsibility to specify
  the analysis design, plan its implementation, and
  supervise its execution but they should also
  enlist the participation of the wider scientific
  community in developing algorithms, writing
  software, and interpreting the results.
• The centers should work closely with LISA
  experimentalists and they should also sponsor and
  encourage related preparatory and follow-up
  observations and studies in other branches of
  astronomy. They should focus on mission-critical
  activities, and should cooperate with and
  encourage independent efforts to supplement and
  check the basic analysis performed by the
  centers. To this end, they should also supervise
  the preparation and release of data and data
  products to the wider community.
• LIST should determine a costing for the data
  analysis activities that is adequate to the task,
  and this should be financed as part of the LISA
  mission. It is desirable to begin some of these
  activities during LISA Phase B, since details of
  hardware design affect the kinds and quality of
  data that can be acquired.

3
 Contents

• Nature and magnitude of LISA data analysis tasks
• Examination of various models for organizing the
  data analysis
• Coupling of Hardware Design (Phase B) and Data
  Analysis
• LISA data centers and their responsibilities

4
1. Nature and Magnitude of LISA Data Analysis
Tasks

• LISA analysis more complex than ground-based,
  clearly critical for mission success. Strong S/N
  and source confusion complicate analysis
• Signals from SMBHs must be measured, removed to
  one part in 104.
• Signals from final plunge and merger phase of
  SMBHs.
• Capture orbits overlap strongly in time. The
  stronger signals obscured by confusion noise
  maybe solve for and remove many signals
  simultaneously.
• Binary systems near the edge of the confusion
  noise cannot be removed individually, but have to
  be solved for simultaneously, perhaps hundreds at
  once.
• Accuracy for binaries improves with time, must be
  updated. Affects all sources.
• Limits on the stochastic background depend on
  removing stronger signals.
• TDI complicates matched filtering, requires S/C
  position, motion information.
• None of these tasks can be performed
  independently of the others. They probably have
  to be iterated several times before the solutions
  converge, and then the solutions need to be
  refined as further data becomes available.

5
2. Examination of Models for Organizing Data
Analysis

• Ground-based gravitational wave detectors do not
  provide a good model for LISA. Better models
  Hipparcos, WMAP, Planck.
• LIGO/GEO data analysis in the LSC 4 groups for 4
  types of sources. LISA does not allow this
  separation. LISA should take advantage of the
  existence of active community, but needs more
  central coordination.
• Important that LISA releases various data
  products for the community to analyze. The data
  set will be small and easily transportable. But
  analysis of LISAs raw data will be challenging,
  so also release cleaned products.
• Hipparcos looked at a many objects and made many
  measurements positions, parallaxes, proper
  motions, binary motions, etc. ESA established 2
  competing teams, set them working several years
  before launch. They tested and validated each
  others methods, codes, and eventual solutions.
• WMAP and Planck set up data analysis teams well
  in advance of launch, at the forefront of new
  analysis techniques and careful statistics.
  Production of the fundamental anisotropy maps and
  statistics is a project responsibility.

6
3. Coupling of Hardware Design (Phase B) Data
Analysis

• The LISA spacecraft and payload design are very
  tightly coupled, so payload data issues need to
  be addressed at Phase B. Needs input from data
  analysis.
• Signal extraction. TDI. Need simulations.
• Housekeeping data. Need to simulate things that
  go wrong.
• On-board processing. Does it need to know
  anything about the eventual data-analysis
  algorithms that will be used? Summary information
  to monitor on-board activity (checksums, other
  consistency checks)?
• Systematics. Simulations, e.g. coupling of annual
  thermal effects to removal of annual amplitude
  modulation. Systematics can affect the apparent
  stationarity (or lack of it) of the instrumental
  noise, and this must be independently estimated.
• Spacecraft control. Effect on data analysis of
  control procedures, such as discharging of the
  test masses, any occasional orbit corrections,
  and even small FEEP events. Also environmental
  disturbances, e.g. large solar flares or the
  nearby passage of a massive asteroid. What
  control/housekeeping data needed?
• Noise estimation. Spectral noise density in LISA
  we have two new problems (1) signal-confusion
  noise and by strong signals that have to be
  subtracted to well in order to estimate the
  noise, and (2) is observed Gaussian noise
  instrumental or a stochastic GW background? TDI
  Sagnac helps at low frequencies, but do we need
  other data at high frequencies for independent
  est.?

7
Phase B (contd)

• Calibration. LISA is self-calibrating, but what
  data is needed to maintain/recover this in
  environmental disturbance or a control event.
• Failures. Simulations ensure that enough data
  exists to recover from possible hardware
  failures. Are there data streams that would not
  be needed in normal operation but which must be
  telemetered or monitored for recovery?
• Some of these issues have already been addressed
  in our existing studies, but rarely in the detail
  that simulations will provide. Any of them could
  potentially affect key data decisions like the
  bit-length of acquired signals and the telemetry
  rate to ground. The determination of how the
  requirements of data analysis influence the
  design of the hardware, control,
  data-acquisition, and telemetry systems demands
  close cooperation between data-analysis experts
  and the spacecraft design team. This in turn
  means that the agencies should begin to fund LISA
  data analysis very soon, so that it has effective
  input into Phase B.

8
4. LISA Data Centers and their Responsibilities

• Nature of the Centers
• There are several reasons for preferring the more
  tightly-organized centers over the looser
  LSC-style collaboration
• Tightly integrated data analysis needs tightly
  integrated development team, working closely with
  experimental team.
• The existing ground-based community is fully
  occupied with analysis for LIGO, GEO, TAMA and
  (soon) VIRGO. Data from Advanced LIGO at just the
  time (2009?) that LISA needs to keep to
  deadlines.
• LISAs data analysis needs to work with full
  reliability from day one.
• The dedicated data analysis centers should be
  funded by and directly responsible to the LISA
  project. There should be centers in both the USA
  and Europe, should cooperate closely, verify each
  other, no need for Hipparcos-style competition.
  Some functions will be exercised by both centers
  for their local communities research funding,
  data releases, user support, outreach.

9
Pre-launch Responsibilities

• Foster the development of necessary data analysis
  techniques.
• Work with LISA hardware designers at Phase B,
  prepare protocols for determining the spectral
  noise density of the TDI streams, for
  validity/quality indicators, and for handling of
  and recovery from hardware failures or errors.
• Enlist university groups to contribute with
  clearly defined responsibilities. Should
  sub-contract and also advise funding agencies on
  awards. LISA center activity should not reduce
  national funding to university groups.
• Enlist astronomy/astrophysics groups in
  performing observations and theoretical studies,
  e.g. studies of BBH galaxies, searches for
  eclipsing WDBs
• Define data products to be released.
• Define the data-analysis pipeline or procedures
  that will produce these products.
• Design user software that will be available to
  the community.
• Develop robust and professionally written
  data-analysis software incorporating data
  analysis algorithms, user-friendly interfaces,
  technical documentation, and user manuals, both
  for the center-based analysis and for external
  users.
• Test and qualify the software using simulated
  data sets.
• Acquire necessary staff and computer hardware.
• Educate the community to use the data analysis
  tools.

10
Post-launch Responsibilities

• Receive, calibrate, process the raw data streams
  into h-data streams.
• Run continuously the data-analysis pipeline.
• Re-run analysis on older data sets periodically
  to obtain improved solutions.
• Produce the agreed data products for the
  community, including real-time alerts.
• Support the university community in interpreting
  the detections and in performing further analysis
  with the data analysis tools designed by the
  centers.
• Support and encourage the astronomy/astrophysics
  community in coordinated follow-up studies from
  LISA observations.
• Foster further research in algorithms needed for
  the study of new kinds of sources that had not
  been anticipated in the original planning cycle,
  and for improving the existing analysis methods
  or correcting flaws.
• Maintain external relations press-releases,
  public education related to the mission and its
  discoveries.

11
Establishment of the Centers

• The list of pre-launch tasks is longer than the
  post-launch list, and some of these tasks must
  begin early (Phase B). Project should begin
  formal organization of data analysis very soon,
  at perhaps a relatively small level, increasing
  when data analysis software is being written.
  Staff numbers probably peak at the time of
  launch. Afterwards fewer programmers, more
  support scientists.
• LIST should estimate required staffing levels
  carefully. Some mechanism should be found to
  allow the centers to support mission-critical
  research in the community, either through their
  own grant-awarding budget or through agreements
  with other funding agencies.  
• Centers could/should(?) be attached to existing
  research establishments.  
• Center funding should be built into the budget
  for LISA. Complex for ESA. Total cost of the two
  centers, integrated over the pre-launch phase and
  a 10-year operational term, could amount to of
  order 5-10 of the total mission budget.
• The success of the entire LISA mission depends,
  ultimately, on the quality of the data analysis,
  and the job is a tricky one. It must be organized
  thoughtfully, initiated in good time, and funded
  adequately. Not to do this would be to subject
  the missions success to an unacceptable risk.