Characterisation Data Model applied to simulated data

1
Characterisation Data Modelapplied to simulated
data

• Mireille Louys, CDS and LSIIT Strasbourg

2
Characterisation metadata

• Should answer the question
• Where, when, what, how precise and reliable
  are the data for one observation?
• FoV, Bandpass, Resolution, Quantum Efficiency,
  etc
• It is a summary of metadata to be used for data
  retrieval as in DAL protocols but also for data
  analysis resampling, source detections,
  multi-wavelength analysis, etc

3
Organising metadata

• Lists the properties of an observation
  coverage, resolution , sampling precision,
  sensitivity, point spread function, transmission
  curve, etc
• These are the quantitative information that we
  can derive from the Provenance (acquisition or
  simulation process).
• Defines characterisation axes as space, time,
  wavelength, observable which is the measured
  quantity like flux, photons, counts, etc
• Categorise them according to a unified framework
• ? Data model expansion capability

4
Physical Axis
Property of the data
Coverage
Resolution
Sampling
Level of description
5
UML model Properties Levels
6
Quality and Errors

• Each assessed property will have the required
  value, unit, ucd fields plus an error on this
  value.
• The typical error on the data, that is the error
  we make when we map sampling elements to
  coordinates, is also needed .
• E.g. astrometric error, photometric error, etc
• They will be attached to the axis on which the
  mapping is done spatial, observable, etc
• Valid for both systematic and statistical errors.

7
Axis description and mapping error
8
Categorising use-cases complexity (1)

• In terms of use cases
• Data discovery and selection level 1-2-3
• Multi regime, multi data type
• Xmatch of metadata to navigate between complex
  datasets cubes, spectra, images, catalogs
• ? Valid for both observed and simulated data
• Advanced data processing level 4
• Physical interpretation, recalibration
• Description of side products to help for data
  interpretation
• PSF variation, transmission curve, quality maps,
  weightmaps, etc
• ? Valid for fine comparison between observed and
  simulated data

9
Characterisation model expansion

• In terms of data content
• More complex axes can be defined
• polarimetry, velocity, visibility
• More data properties can be added
• In terms of dependencies
• Coupling of characterisation axes
• expressed as functions
• e.g. Resolutionf(pos,em,time)
• expressed as variability maps, e.g. PSF
  variations maps

10
PHENOMENON
PROVENANCE
DATA CHARACTERISATION
Scientific knowledge
11
PHENOMENON
PROVENANCE
CHARACTERISATION DM
Scientific knowledge
LEVEL 4
Interpretation metadata
Error maps
Observation Process
Object of interest
Input Data
Transmission curves Weighting functions PSF
variability
Proposal
Ambiant conditions
Filters
Instruments
Output Data
LEVELS 1-2-3
LEVEL 4
12
Simulated vs observed data commonalities and
differences

• They share axes, data properties, format (?),
  coding.
• spatial axis observed data are always centered
  on real sky position, simulated data are not
  necessarily.
• Calibration information is extracted from the
  Provenance information
• Specific interpretation metadata can be generated
  by the simulation computation.
• Maps, multivariable functions to be described by
  level4 classes in Characterisation.

13
Conclusion

• The Characterisation model can carry out the
  description of simulation output data.
• Version 1.0 currently to describe the top 3
  levels the footprint of the data on the axes
• The next version of the model will emphasize the
  level 4 structures.
• Simulation codes to be described
• DALIA, Frederic Boone, Obs. Paris, LERMA
• An homogeneous dynamical interface for various
  simulation codes XML schema description
• Compatible to workflows in data processing
• To be part of the Provenance DM effort?
• Need for Phenomenon modeling