4D NetCDF Rendering

[what_nick]

Hi All,

This thread considers the issue of rendering data with Z-component particularly in curvilinear and sigma grids. Unstructured grids such as hexagonal fuller grids will need to be considered later as this project - https://github.com/asascience-open/NetCDF-Java-UGRID develops.

I have done some work along the lines Jim Miller has done to render NetCDF in 3D using glyphs as opposed to textures. This is especially useful for vectors and tensors. Here are a couple of sample videos:

http://dl.dropbox.com/u/15821898/Cyclone_Yasi_Atmos_Ocean.mov
(Ocean and Atmospheric model of Cyclone YASI - Reanalysis i.e. post-facto)

http://dl.dropbox.com/u/15821898/Curvilinear_Grid_NetCDF.mov
(Curvilinear grid with the vertical levels being a sigma grid)

The source for the above videos can be found here: https://github.com/whatnick/glob3_csiro . Randy holds the relevant data backing the videos.

It will also be useful to integrate Jim's Runge-Kutta integrals to produce the streamlines from seeds: http://commons.apache.org/math/apidocs/org/apache/commons/math/ode/nonstiff/RungeKuttaIntegrator.html .

These tasks would be some of the things for the ESA SOCIS student to consider and integrate into iGlobe.

Cheers,

whatnick.

[what_nick]

Alex on the ESA project has indicated an interest to implement the Runge-Kutta integrals in windfields. There is some sample U-V data (u_wind,v_wind and others) in the Alaska Thredds server we can use for this purpose.

Cheers,

whatnick.

[jimmiller]

I have to admit that I don't know what a "sigma grid" is. Can you enlighten me? Also, let me know if you want/need any help with the Runge-Kutta integration.

[what_nick]

Hi Jim,

Here are some notes on the Atmosphere Sigma Co-ordinates from the CF convention page - http://cf-pcmdi.llnl.gov/documents/cf-conventions/1.0/cf-conventions.html#id2897716 . NetCDF follows physics more closely than computational ease.

I was considering using commons-math implementation of Runge-Kutta, since it is supposedly well tested and has a clear licence. This also brings about the rules for use of third-party libraries in iGlobe and licence considerations. At CSIRO we have the following rule of thumb:

Don't touch GPL or fancy proprietary licences, opt first for things under MIT or BSD, we can also work with Apache or LGPL.

Let me know your thoughts on this, we might possibly start a separate thread along these lines.

Cheers,

whatnick.

[jimmiller]

The Runge-Kutta code I eventually used in earnest was a translated/adapted version from Numerical Recipes. Gray area, but certainly was derived from their code base. I agree in general - this issue needs pretty clear rules/guidelines and careful monitoring.

[what_nick]

I was investigating slice based rendering as we are currently doing as opposed to glyph based rendering which is good for Tensors. I came across a demo (http://www.derschmale.com/demo/volumeRendering/ultra-lq-switch.html) which illustrates much better what I intend with slice based rendering than I can put in words (http://www.derschmale.com/2009/07/20/slice-based-volume-rendering-using-pixel-bender/). Feel free to verbalise.

Cheers,

whatnick.

[jimmiller]

Yes - this is a nice demo. There are a variety of volumetric rendering schemes like this that are applicable to scalar data defined on a 3D grid (or continuously in 3D). Some general categories are ray tracing, contouring (e.g., marching cubes), numerical integration, or slicing (in which a series of planes - e.g., perpendicular to the line of sight - are created, used to sample the field, and then rendered). I didn't read the paper yet, but a couple of the labels I read lead me to believe it may be similar to this latter scheme. Many of the methods are easily mapped to GPU programs (using CUDA or OpenCL, for example) to get interactive performance.

[what_nick]

Hi All,

Uwe has some preliminary scalar field rendering going with slices at different z-levels for the Cyclone Yasi data. Here is the preview video.

http://www.flickr.com/photos/whatnick/6237082348/in/photostream

Enjoy,

whatnick.

[michaelwolf]

HI All,

Why the latest version of iGlobe cannot display the nc data when checked the leaf checkbox? but when I select "Play" button of animationcontrollayer, the scalarsurface layer will be visible.

Thanks,

Michaelwolf.