Notes from meeting with Scholin, Haddock 3/9/05

What is motivation?

comment with platforms, IAG

CS:  Look at broader science themes that are not well developed, rather than individually. (KG: first couple of days of offiste are for that. These meetings are litmus test.)  A way of organizing ourselves.  Stop with the fluff, too all-encompassing.  Derive functional requirements for the major science themes we want to tackle.

Number of not well developed themes, science has to develop them. (Some documents exist, MSE.

Among them: Notion of event response probably drives a lot of what he's doing.  Case of ESP, low freq high bandwidth instrument.  Only one physical sample with lots of tests performed, very resource constrained. Depend on context. Developed words toward that for Tom's proposal.   

Notion of interpolating between data points becomes quite important, make guesses about what's in between.  Require physical samples.  AUV response will use a nowcast style model.

SH AOSN (writing workshop synthesizing papers), could take a lot more advantage of continual high-res models, CODAR, circulation models can be used to trace events through time.  Put it into context, connects with regional collaborators.  Yi Chao.  Nested grids with finest on Monterey Bay.  Gearing up not optimal solution, better to solidify relationships.  Their LAS server supports generating pieces on request, e.g. for Monterey Bay.  LOOKING another opportunity to keep that going.  Almost like NWS is larger pot of money to keep going.

SH: Could take advantage of event response.  Part of AUV allocation is flexible days for intense sampling or outside scheduled window.  Could put biolum sensor on mooring for detection, one available; other related triggers (rain, wind), could be tied back into models.

Science strategic planning? Started up around eng reorg.  Up until now single PI, collaborations are ad-hoc. Ken J would like to see us look at ourselves more thematically.

CS: Notion of being more interactive with your science (lab in a can, lab on a mooring).  Rather than continuous instrumentation, start interacting with in situ instruments. Capacity is talked about a lot, but not specified how.  That's where high bandwidth, real time comm come into play.  Crude but widespread sensor net could generate lots of information, allow detection in instrument or externally that would drive further study of measurement. What's the key thing that'd make that work, that you don't have today?  ...Must keep good records as you go along, value lost without that metadata, have to know precisely when. Remote in situ lab, only have limited options.  Also decide when to archive or not.

SH: Virtual notebook for all the data that is collected: physical context, call in all the associated information.  Instead of paging back through notebook looking for everything he knows.  GUI that ties in to samples, VARS databases, take on ships.  Don't have to be on network to interact.  Better the front end, the more he interacts.

CS: A key piece of metadata is the age of the sample.  After certain time, opportunity is lost (characteristic is labile).  

Not necessarily lots of labs, but ability to interact remotely with just one.  3 to 10 ESPs for 4 months period, like LEO-15 (turn it on and off), limited duration and focus.  

Need something like Steve suggested?  Yes, Rich's MDB is very much like that, capturing the local points, can link to photos and notes, generate reports.  Licensing issues put off people (energy of activation barrier).  

How dependent on contextual information? Somewhat, but resource-limited (personnel and data sets).  CS: Chris Preston's microbial work provided context from M0 and CODAR synopsis, satellites when they can. Deploy on schedule in both cases.  During analysis the context is critical, may boil down to one image.

SH: Video and image processing (transect type info from normal dives) would be great to have for context.  Great to have high throughput of associative data.  Data quality is a problem, data integration (currents, temp) would be valuable to put together with the specific observations.  

CS: What are themes cutting across various instruments?  Fluidics, etc.  With respect to image analysis, expertise cuts across and would be good to be able to spply to multiple projects. This architecture speaks to multiple applications.  [Note: This is a growing theme on the technical side.]  What kind of expertise is needed within a group?

CS: Their style is to use whatever tools are available, doing ad-hoc analysis.  So there isn't a particular visualization need.  Still too immature.  Lab in a can is much more pro-active.  No preconception for an interface.  4D will be important to them.  The point measurement is well obtained, but contextual features are not available.  Profiling, other data would be real helpful.  Time resolution is very important.  

CS: Their logged sample doesn't go anywhere else to provide context for other measurements, either.

SH:  You see what you see, but want to extract contextual information?  Imagine a nice visualization of data with overlaid observations.  Past discussions have been about AUV sections over time, putting together X-Y surface plane with 3/4D data.  Trying to put together all the features in a way to understand it.  VIS5D/IDV are possible solutions here.

What about the on-the-fly visualization/goal?  Useful future capability.  SH focusing on after the fact analysis. Seeing two individuals of same species, figuring out how to investigate association.  "Little red light on the boat."  Wants TiVO rewind on the video, replay both image and context data.  Icon mapping to annotated entities and displayed next to vertical transect display.  Being able to connect annotated value to video frame grab would be good.

Discussion of new thrusts vs continued upgrade of existing systems.  Need to integrate priorities of the two when allocating resources.  Include users in the beta testing of systems to make sure system is beta tested before deployment.

Comment/problem reporting capabilities within software would be good.


John