[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Fwd: netCDF and IDV for the Deepwater Horizon

CJ Beegle-Krause wrote:
Hi Jeff,

Ansley Manke is working on some example files as you know.

When I read the standard, I'm not sure what to do with particles. Unidata and Hazmat - we will have to be careful of how we use the word "trajectory" because the word has more meanings over here. We run trajectory simulations with 1000 to 10000 particles, and keep the particle history: location, time, mass, etc. A trajectory map is a time slice of the particles processed to map a particular type of map. We have more than one type of particles. For example "best estimate" and "uncertainty". We use game theory concepts: best estimate is where we think the oil will mode likely go over the forecast period. Uncertainty is sets a bound on other possible trajectories - we post process those particles differently to create this error bound on the map. Another dimension is what the particle is, such as "oil", "copepod" or something else. Right now we're just running oil particle that have different diameters.

Things can happen to a particle along an individual particle's path, such as changing mass, and sometime we may do this as post processing.

We have MOSS files as an ASTM standard for our particle trajectories, but they are text and more useful for GIS than for modeling. Products currently are snapshots in time. With 4D capability, we need to rethink how to advise people for subsurface. Mostly what we will need to do is help 8 vessels know the day before, where to sample the next day to find what they are looking for. The other operational decision is whether or not to use subsurface dispersant applications: "go" / "no go". Here is where we really want to be able to overlay models, observations, climatological T,S, etc. into an enviroment that can move through time.



Hi CJ:

To clarify, my suggestions are about helping your group put your data in a standard netcdf format. This will take at least a few weeks, probably much longer, so is not any kind of short term solution.
This would allow you to use IDV and other standard packages to do 
visualizations. I would guess that your operational tools would need to be a 
mixture of general tools like the IDV and special purpose ones that your group 
would continue to develop.

In these files, you can use whatever vocabulary your community wants. You have to add 
some CF specific metadata following CF vocabulary, but that is more or less "under 
the hood".

If you want to proceed, we can help with specific examples. We are interested in making 
sure the new CF standard covers as many use cases as possible. In the above, I guess you 
would use one of the variants in section 9.4. Each particle would be a 
"trajectory" in CF lingo. You would probably store all data for a simulation in 
one file. The idea is to make sure this is efficient for downstream processing.