Subsetting to save time and bandwidth

March 25, 2016
Posted by: Dave Stepaniak

Note: This page was originally sourced from our Blogger page: http://ncarrda.blogspot.com/2016/02/subsetting-to-save-time-and-bandwidth.html

Many people use RDA data to initialize the WRF model.  Some researchers ask for help to cope with bandwidth constraints.

Others may not even be aware that they are running into bandwidth constraints until too late--when they get the "429 Too Many Requests" error after their IP address has been blocked for overuse of RDA resources.

We want to help you get the data you need to do your work.   But you will save time and bandwidth if you keep a few simple points in mind.


WRF does the dynamical downscaling for you.  If you don't see any significant difference in WRF forecast quality with between the 0.25 and 1.0 degree GRIB input files, then you need only 1/16th the bandwidth!  But, if your domain includes complex terrain, you will likely see better results with 0.25 degree input.

Use nested domains to downscale iteratively to the desired resolution.  See the WRF Tutorial Nesting slide deck.  You can use 1.0 degree data to initiate an 81 km 'parent' domain with 27 km and 9 km 'child' and 'grandchild' nested domains.  You might use 0.25 degree data to initialize 27, 9, 3 km domains and nests.

WRF nested grids.
There is no reason to download a complete global grid to run a regional model. WRF requires only enough input data to cover the desired domain plus a 'relaxation' zone around it. A10-degree border is more than sufficient.

Specified and relaxation zones (yellow/blue) of 5 WRF parent domain (green) grid points.
For example, if running a WRF simulation of a region 30 degrees tall and 40 degrees wide, you can safely supply only input data 50 degrees tall by 60 degrees wide.

(50*60)/(180*360) = 0.0463

Downloading only the subset of the global grid that you actually need saves you more than 95% of the bandwidth of transferring the entire GRIB file!  You and your colleagues sharing the same IP address will be done 95% faster, reducing the number of concurrent data transfers on your local network.

[Aside: Check your download manager to request no more than 10 concurrent downloads at a time per user.  If you don't, you will get the "429 Too Many Requests" error.]

Convinced?

Use the 'Get a Subset' option.   As we say in Colorado, work faster and ski more!

If you are using a script instead of the web interface, note that most of our GRIB data sets support the OPeNDAP protocol. See these NCDC OPeNDAP URL syntax examples for formulating your own scripted requests.