AnsweredAssumed Answered

sphere for WRF data in Lambert Conformal Conic?

Question asked by dca.uw on Feb 14, 2018

Hi everyone, I have some WRF model output data in NetCDF format, and I would appreciate any guidance on projecting it from anyone with experience with data from this model. (WRF = Weather Research Forecasting)


These WRF runs use a Lambert Conformal Conic projection with a center of -121, standard parallels of 30 & 60 degrees North, and a latitude of origin of 45 degrees North. I think I have that figured out, but my data will be off if I don't use the right sphere/spheroid. My understanding is that WRF uses a sphere of 6370 KM with no flattening. Does this look right? 




If that is right (or if I need to adjust it), then I should be all set. I would appreciate any feedback I can get from anyone with experience in this. Based on slide 20 of this presentation, it looks right, but I'd appreciate as much help as I can get!


That said, my process to bring these into ArcGIS is not for analysis is not straightforward at all, and I'd appreciate any suggestions to improve it. Here is how I do it: 

  • Use Make NetCDF Raster Layer to import to an in-memory raster and save to a GeoTIF.
  • The GeoTIF has headers for the coordinate system, but it comes in as pixel coordinate space with no registration. Unfortunately, the Esri Define Projection did not work on this. I saw several other people griping about this in this forum. 
  • I use gdal_translate.exe -co PROFILE=BASELINE to strip out the file's coordinate system headers in the GeoTIF. 
  • I pre-generate a TFW and copy it to <filename>.tfw in the folder where <filename>.tif lives. (I do this for each time slice in the NetCDF that I export.) The NetCDF gives me the lower left, and I need the upper left in the TFW, so some basic math is needed with the cell size and number of rows. 
  • Use Define Projection to put it into the right coordinate system. It works now that there are no GeoTIF headers and a TFW file exists. 


I'm doing this in Python and it's automated. It doesn't take long, but YUCK! It seems like a lot of steps to get the data into the right coordinate system, especially considering that it came from a file where the coordinate system is well-defined, albeit in a different format. After this, I do some resampling, averaging, and export to polygons, so I feel like TIFs are a good intermediate format, but there is clearly room for improvement...