Jan Mandel/Blog/2010 Oct Nov
These are mostly notes from our regular meetings. See also
- WRF-Fire commit graph (scroll right to see the dates and authors)
- list of all WRF-Fire pages on openwfm.org
- WRF-Fire development notes
October 4, 2010
WRF-Fire: FireFlux movies
- Bedrich produced a number of nice movies from FireFlux simulation using VAPOR: File:W-sideview.mov , File:P.mov .
The W-sideview.mov is the DVR visualization of the vertical component of the wind and it's colormap (from a different viewpoint) can be seen here: . Blue/violet colors mean that the wind is blowing up, yellow/green means down. The opacity for values around zero has been decreased so the velocities around zero are transparent.
CCM Colloquium: The Fourmile Canyon fire
- The movies used in my presentation The Fourmile Canyon fire, Google Earth, and the WRF-Fire simulator
- Jon has figured out and wrote great code for Visualization in Google Earth both for EpiSim and WRF-Fire.
CVS repository daseminar
- I have set up a new CVS repository for papers and such, see instructions how to contribute. The repository is exported to the web hourly. The old dddas03 repository is being retired and it will soon be read only. I will select and transfer relevant files and directories. Please let me know if you want something copied rather than just commit a copy of an old file, because I can copy files including the history.
- Do we want subdirectories in references for papers in subject areas: fire, epidemics, data assimilation. Why not? Can soft link papers that fall in more than one category.
- Ashok's way of naming fulltext files by author title vs mine by bibtex key Name-YYYY-ABC. What is better?
- Loren proposed making entry on the wiki and link to fulltext.
- Why not generate links from the bibtex files? Jon generated bibtex links years ago, would need some scripting to add clickable links to pdf on the html version of the bibtex file. See also How to write BibTeX files.
October 6, 2010
- Janice objects to me listing myself as leading the WRF-Fire software development. We'll need to figure out what the structure is. Also in the same edit, I'd like to get Kara's code mentioned with the improvement to ignition.
- Jon has added more to Visualization in Google Earth and I have copied the page to openwfm:Visualization in Google Earth with redirect from openwfm:How to visualize WRF-Fire output in Google Earth.
- Volodymyr cleaned up some the fuel code, standalone tests OK but now getting NaNs in gfortran though not pgf90. Maybe this merge with master is to blame. Still needs to remove prints. To verify the problem, printed after fuel_left, got fuel_frac_end(410,198)=NaN.
- Try setting in fuel_left fuel_frac(410,198) the value early on and see if it changes.
- Find the exact commit where it happens first.
- Try valgrind to catch access to uninitialized memory.
- Ignition: Replaying history from ignition times as level set function done two weeks ago but I have made no progress since then, need to think more.
- Need to add reading the ignition times from a file in ideal.exe
- Tool to convert a fire perimeter in a shapefile to ignition times/level set function - the next project for Volodymyr.
- Jon added lfn_hist to wrfinput as 3D as I asked... but turns out I need only 2D, embarassing. Still trying to figure out how to take advantage of the extra dimension - several igntions?
- Ignition can be of 2 types: advancing the fire during the ignition, or not. When replaying history, clearly not. But how about taking advantage of the extra dimension to do gradual ignition?
- It may be easier to do gradual ignition by imposing ignition times/level set function in a variant of existing code.
- Other things to do in the coming week:
- Jon - continue on Meadow Creek fire, document for wrfhelp
- Bedrich - annotate the movie files, include jpegs, explain colors; see about something as in Norton-2009-AVH. Save VAPOR sessions. All pictures and movies will need to be redone when Adam finalizes the FireFlux simulation.
October 7, 2010
- Added Bedrich's movies   to our YouTube channel
- Found Craig's Clements' channel with the FireFlux movie and more.
- Our previous channel is still there, how to join them?
- Nina Dobrinkova   is joining us in the WRF-Fire group at UCD for a month today. Some of her papers: Dobrinkova-2008-GMF Dobrinkova-2009-PSE Dobrinkova-2010-WAB
- Created How to get started in the fire group
- Start by running the existing simulation here
- Nina's presentation File:N Dobrinkova.ppt
- Needs to put actual numbers for fuel, adjust fuel properties for correct spread rate, see How to diagnose fuel properties in WRF-Fire
- Atmosphere resolution 160x160, fuel by sectors in database screen, each sector a polygon on a map.
- Fuel is in met_em files also 160x160, how were they created?
- Running in real, but ignition a thick line, fuel artificial by altitude
- Fire spread rates
- Examining and changing WRF files in Matlab
- Start matlab in WRFV3/test/em_fire to set up the search path right
- Then cd where the files are and use:
ncload wrfinput_d01to read all variables from file (e.g. wrfinput_d01) into matlab workspace, and something like
ncreplace('wrfinput_d01','ZSF',zsf)to replace a variable in the file
October 18, 2010
Data assimilation seminar
- Cholera epidemic in Haiti
- Jon reported instability also in the Harmanli fire. I have now rerun the case and eventually got vertical cfl violation at one corner of domain , consistent with his result. Is this perhaps an outflow boundary condition and Euler equations issue?
- Debugging fuel integration: violation of the condition that fuel fraction should be between 0 and 1, is 1.02. Is this just rounding error? To find out, try computation inside the subroutine in double precision. Try to adjust the quadrature of the exponential function - is the crossover to Taylor explansion set correctly for single precision?
October 28, 2010
NCAR MMM seminar: Aime Fournier
- Aime Fournier, Development of Wavelet Methodology for WRF Data Assimilation. pdf earlier version
- Main idea: estimate covariance in wavelet space and discard off-diagonal terms, much like FFT EnKF.
November 8, 2010
- Jonathan Beezley, Local versions of the ensemble Kalman filter
- Jeffrey L. Anderson. An ensemble adjustment Kalman filter for data assimilation. Monthly Weather Review, 129:2884–2903, 2001. local link
- Jeffrey L. Anderson. A local least squares framework for ensemble filtering. Monthly Weather Review, 131:634–642, 2003. local link
- Edward Ott, Brian R. Hunt, Istvan Szunyogh, Aleksey V. Zimin, Eric J. Kostelich, Matteo Corazza, Eugenia Kalnay, D.J. Patil, and James A. Yorke. A local ensemble Kalman filter for atmospheric data assimilation. Tellus, 56A:415–428, 2004. local link
- Pavel Sakov and Laurent Bertino. Relation between two common localisation methods for the EnKF. Computational Geosciences, 1-13, 2010. local link
November 11, 2010
NCAR MMM Seminar: Yann Michel
- Yann Michel, The Use of Displacements in Data Assimilation. pdf slides File:NCAR seminar 20101111 ym v1-2.pdf local copy. Uses morphing from Beezley-2009-MEK.
November 15, 2010
Data assimilation seminar: Aimé Fournier
- Aimé Fournier, Development of Wavelet Methodology for WRF Data Assimilation pdf
- Further info: WRF Data Assimilation System Users Page
- Background on wavelets
- Gil Strang, Wavelets, American Scientist 82 (1994) 250--255
- Aime Fournier, Introduction to orthonormal wavelet analysis with shift invariance: Application to observed atmospheric blocking spatial structure. Journal of the Atmospheric Sciences, 57(23):3856–3880, 2000. local link
- Numerical Recipes in FORTRAN 77, 1994, chapter 13.10 Wavelet Transforms local link
- How to use Wavelab
November 29, 2010
Data assimilation seminar: Loren Cobb
- Bayesian Tracking on Riemannian Manifolds
- Background: Monte Carlo Tracking on the Riemannian Manifold of Multivariate Normal Distributions, by Snoussi & Richard, IEEE DSP-2009, 04-07, January 2009 (invited paper).
- Main idea: When the model restricts the data to a differentiable manifold, then statistics can be defined intrinsically (on the surface of the manifold) rather than in the usual extrinsic form (within the larger space in which the manifold is embedded).
- Essential concept: Geodesic distances on the manifold can be defined via the manifold's family of inner products (the tangent space at each point on the manifold has its own inner product).
- Recent results suggest that all of classical statistical estimation can be expressed intrinsically on smooth Riemannian manifolds.
- Optimization, e.g. the method of steepest descent, can also be defined on Riemannian manifolds.
- "Simple" application: Kalman tracking of an object in elliptical orbit.
- Interesting application: Tracking a covariance matrix on the manifold of positive definite matrices, using a particle filter.