In this I want to figure out how astronomers work - what do they do in order to solve a given problem. I'll list what I think might be a reasonable series of steps and then ask the project astronomers to amend and comment on this.

I'm mainly interested in how astronomers will use the VO (assuming the sort of infrastructure that AG has implemented) but, given that the VO will not do everything, need to consider non-VO activities.

I think this list can serve several purposes:

• for myself, I want to look at ways in which semantic web developments might apply to the astronomer's working practice;
• for the engineers, so that they can better understand what astronomers do and possibly come up with additions to the workbench and components;
• to compare the activities to the support provided by some of the VREs (virtual research environments) under development within various JISC programmes (e.g. http://www.jisc.ac.uk/index.cfm?name=programme_vre).

additions by AnitaRichards 01 Aug 2006 - delete at will, or deitalicise if acceptable

Can all our existing parameterised workflows be described like this?

Solving a problem

• state problem
• analyse problem
• define workflow

• find dataset
  • Multiple data sets are often required for comparison, to cover the full area etc. The data selection process may involve choice of named data sets, or maybe just specification of regions and wavebands or times, plus optimum resolution and sensitivity ranges.
  • Some or all data used may be simulated, so "find data" would be replaced by "select tool to generate data"
• select trial data
  • might be done interactively or left to a VO algorithm to decide what data best matched selection criteria.
• check data
  • This may include finding and applying data processing, calibration etc. e.g. to match resolution or convert photometry
  • May be partly or fully automated using proper data models if software is available (either provided by VO or by interface to data provider facilities)
• find function (s) (application, library function etc)
• apply function (s) to trial data
• check trial results
  • Visualisation
  • Statistical tests
  • Other diagnostics for automatic checking e.g. allowed ranges of values

• select full data
• check data
• apply function (s) to full data
• check full results

• add data selection to workflow
• add function (s) to workflow
• run workflow
• check results
• Document processes used (data, methods...) - so these need to be transparent at the level of detail needed for a typical paper or thesis
• Produce summaries of results (figures, tables etc.) for discussion and publication.

Cycles

The above steps are not really a list but a set of cycles. Each time something is done, the astronomer will check it and, if the result looks invalid, will cycle back to change something. In some cases the user could incorporate conditional or checking algorithms. I often find it easier to run entire pipelines which print out diagnostic tables or plots at decision points, and check those at the end at leisure, rather than wait on separate stages.

Often, different steps will be performed days or even months apart if the user is busy or is waiting for an unavailable service (especially if they ar dipping in and out of VO tools). So they need to be able to establish easily where they have got to and carry on.

Day to day work

What things do astronomers do in their day-to-day work that does not involve solving some astronomical problem? I would reword this as "apart from desk/laptop-based personal research" but in fact many of the activities below do involve the VO e.g. getting data to plot for a lecture slide, or demonstrating the use of VO tools as part of planning an observation etc..

• Teach/support scientific visitors/observers/answer questions
  • Including preparation, demonstrations, exam marking
• Prepare observing proposals
• Make observations, reduce data,(for self or others).
• Write/design software, pipelines, documentation.
• read journal papers
• write papers and talks
• Collaborate (talk, visit, email, phone...)
• apply for grants
  • contribute to other grant applications
• check status of projects
• Go to conferences (and prepare talks/posters/demos etc.)
• Go to admin/non-specific meetings e.g. departmental, conference planning
• Admin (e.g. form-filling for RAE, expenses...)

Issues

We should remember that most astronomers/solar scientists (hereafter users) will dip in and out of AstroGrid to use the bits they like, and in practice most people do not start by using AstroGrid for a full workflow. Even when they do, there will still be some breaks either to use their favourite software in an old familiar interface (even if they could use a VO tool!), or to test some new private data or code, or to tweak something interactively, etc. I think that it is only incorporating software into the VO if either there is a big demand for convenience (use in workflows) or if we can add value by replacing some of the manual decision making by parameter values automatically derived from metadata. So maybe a less megalomaniac distinction would be Research (including use of VO facilities) v. non-research.