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

[IDV #YZO-547503]: data ingestion into IDV, plotting vectors, display problems



Greetings Kemal!

> Hello,
> 
> I am contacting you to report several issues of IDV that I am experiencing, 
> so I can get some help. I created a tar file that includes all the figures 
> and input data with one time step only to help you to 
> regenerate the issues I am mentioning below and placed it onto:
> 
> ftp://eos.arb.ca.gov/pub/outgoing/kgurer/idv_help.tar.gz
> 
> I am trying to plot wind vectors near the surface from MCIP output of WRF 
> model result and WRF output into IDV3.0u2 on a 64-bit Windows platform having 
> 4 Gb of RAM (HP E8400, 3 GHz, Core2 duo). I 
> can ingest WRF output into IDV using any one of three file type options 
> available in IDV: 1) "aggregate WRF netcdf grids by Time", 2) "Grid files 
> (netcdf/GRIB/OpeNDAP/GEMPAK), and 3) aggregate grids by 
> time, and the wind vector fields look very similar (see figure 
> "feb_2_12Z_from_wrfout_u_v.jpg).
> 
> Because of the use of different grid systems/projections in WRF and MCIP, the 
> least complicated way to compare both outputs is to plot wind vectors using u 
> and v components at 10 m height. These 
> variables are U10 and V10 in WRF, and UWIND10 and VWIND10 in MCIP's output 
> called METDOT3D.
> 
> 
> 1)      When I compare the two plots (WRF and MCIP input) using IDV, I don't 
> see any resemblance of the two fields to each other at all (see figures 
> "feb_2_12Z_from_wrfout_u10_v10.jpg" and 
> "feb_2_12Z_from_metdot3d_uwind_vwind.jpg).

I was able to load in the metdot3d data and get a vector field very similar to 
the WRF field. However, if I flipped the u and v inputs into the flow field 
formula, I was able to generate a plot like 
"feb_2_12Z_from_metdot3d_uwind_vwind.jpg" - double check to make sure you are 
entering the wind components correctly.

> When I independently plot wind vectors using NCL on my Linux box from WRF and 
> MCIP, I see that both plots are identical. Then, this raises the possibility 
> that I am not reading MCIP output (METDOT3D) 
> into IDV correctly since vector plot of WRF output on windows is the same as 
> the one generated on Linux box (The Linux cluster doesn't have a good 
> graphics card, and as a result displaying vectors takes 
> very long time).  What type of input data format I should choose from the 
> IDV's list to ingest I/O API MCIP output properly? 

I used the file type ""Grid files (netcdf/GRIB/OpeNDAP/GEMPAK)". However, there 
is one problem with the data file - the vertical coordinate "Lay" isn't 
properly formatted, so the 15 layers are assumed to be in order and are given 
evenly spaced coordinates between 0 and 1 with no unit. This can be fixed using 
NCML, but I'd need some more info about the structure of the vertical grid.

> I came across with the following reference to ingest I/O API netcdf formatted 
> files:
> 
> http://www.cmaq-model.org/cmaqwiki/index.php?title=CMAQ_version_5.0_%28February_2010_release%29_OGD&CFID=3863765&CFTOKEN=92137774#Integrated_Data_Viewer_.28IDV.29
> 
> (see section 12.8)
> 
> However, this information doesn't help me in choosing the right format in IDV 
> to use MCIP output. Is it not available in IDV yet? If the proper format 
> available in IDV, what is the name?
> 
> 
> 2)      When I plot wind vectors near the surface using 3D variables u and v 
> from wrfout, the vectors over land is almost completely empty while they 
> exist off the coast. (See figure 
> feb_2_12Z_from_wrfout_u_v.jpg). This first suggest that terrain height might 
> be obscuring the wind vector over the land but sporadic appearance of wind 
> vectors over land suggests that something else is 
> going on. I look at the command window where IDV is launched to see if there 
> are any error messages, but I don't see any. IDV uses 1.2Gb of available 1.6 
> Gb of memory, so it should be enough; do you 
> agree? What  do you think is happening?

I see the same thing. I think this is an issue with terrain, because as you go 
to higher levels the 'gaps' begin to fill. However, I do have one question - 
did you destagger your wrf output before loading it into the IDV? It does not 
look like it, but I just wanted to be sure. If not, I would recommend that you 
do so - the performance of the IDV will definitely improve :-)

> 
> 3)      When I plot the wind vector field from one data file and then plot 
> the another vector field from the other data file, the vector size in the 
> second one is plotted short, so the vectors are not visible. When 
> I increase the size of vectors from 4 to say 30 to see it better, the vector 
> size gets smaller in the first plot. When I go back to the first plot, this 
> time same situation happens in the second one. This goes 
> back and forth. I cannot keep the both fields' vector sizes at visible 
> ranges. How do I eliminate this?

This sounds like a potential bug - I'll have to dig into this more.

> 
> 4)      I tried to find the graphics card information to help you to locate 
> the display problem, but I couldn't. Only information I have on my 
> workstation is it is an HP dual core with two 3 GHz processors (total 
> or 4 cpus) with 4 Gb of RAM. I didn't see an option to change the display 
> capability to OpenGL as the research to improve the results suggested. Also, 
> changing the display capability from 32bit to 16 bit 
> didn't help.

In order to change the java3d backend to openGL, you need to edit the 
runIDV.bat file in the IDV program folder. However, I think openGL is used by 
default (check the comments in the runIDV.bat file confirm this...I don't use 
Windows often, so I can't offer much help there - but I do know the information 
regarding openGL / Direct3D is in runIDV.bat.).

Cheers!

Sean

> 
> I appreciate any help that you may give. Thank you very much in advance. With 
> regards,
> 
> Kemal.
> 
> 
> 


Ticket Details
===================
Ticket ID: YZO-547503
Department: Support IDV
Priority: Normal
Status: Open