Wind profiles used by predictor

[Mark Conner]

Gents,

I'm curious what wind levels your GFS model data pulls give you especially for winds above 100 mb.  I think there are several flavors of GFS data that can be pulled from NCEP, some of them more detailed than others.  We've found here in the central US (esp. March-October) that if the vertical detail is too coarse it can lead to more errors in the landing prediction.

When I used some of the older tools like the EOSS Balloon Track and Near Space Venture's site, I would hand edit the wind fields above 100 mb, either using actual wind sounding data (generally OK above 100 mb for < 48 hours out) or merging multiple sources of wind data.  How difficult would it be to (1) present the wind profile as a table, and (2) offer the user the ability to edit it (add/delete rows and modify values)?  I'm a meteorologist, so I get into messing with details like this.  :)

Thanks and 73,

Mark N9XTN

[Adam Greig]

Hi Mark,

We use the files like `gfs.t12z.pgrb2.0p25.f090` from http://nomads.ncep.noaa.gov/pub/data/nccf/com/gfs/prod/gfs.2016041512, which we've found to be the most detailed available, and generally seem to give good results even for high altitudes. It is already an ensemble forecast so should be about as good as you get from merging other data sources or incorporating sounding data, but I'm definitely not a meteorologist so maybe that's not quite how it works!

It's not really feasible to present or allow editing that data - we interpolate the forecast in time, pressure-altitude and position around the current balloon position so there are actually 16 neighbouring wind velocities being considered for each timestep of the solver. We found this interpolation gives much better results than just using the single nearest datapoint in time/altitude/position. Plus, the data is stored and accessed from a huge memory mapped array on disk that we keep up to date with the latest forecasts, which is a fixed resolution grid. Adding/deleting rows would be a real challenge and even allowing editing on a per-prediction basis quite tricky.

The code is all open source if you'd like to have a look; the newest version is https://github.com/cuspaceflight/tawhiri/ and you are welcome to try messing with it yourself!

Cheers,

Adam AD6AM

--
You received this message because you are subscribed to the Google Groups "UKHAS Tools" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ukhas-tools...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Adam Greig]

Oh, and to be a little more specified about pressure levels, we get the following above 100mb:

1, 2, 3, 5, 7 from pgrb2bf and 10, 20, 30, 50, 70 from pgrb2f.

There are some more details about how we deal with the dataset specifically here:
https://github.com/cuspaceflight/tawhiri/blob/master/tawhiri/dataset.py

and the interpolation of it here:
https://github.com/cuspaceflight/tawhiri/blob/master/tawhiri/interpolate.pyx

Adam

[Mark Conner]

Adam,

Thanks for the background.  That is probably the best data exposed to a public server.  The BUFR wind profiles run for specific sites have more detail, but those may extract from the raw model data available only within NCEP.  However, those profiles would be hard to use on the CUSF site since they are for limited points.

I see in your Python code that you get 25-mb increments below 100 mb, another good thing.  Also, I imagine your treatment of the data via interpolation is probably superior to hand editing, esp. given the level of input detail you use.

You guys do a great job with the predictor, and I've had pretty good results with it.  On the last flight I participated in, the actual landing was about 1 km away from the morning prediction.  Crossing my fingers that my flight tomorrow turns out the same.

73 de Mark N9XTN

You received this message because you are subscribed to a topic in the Google Groups "UKHAS Tools" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/ukhas-tools/FIaT4HSDCtg/unsubscribe.
To unsubscribe from this group and all its topics, send an email to ukhas-tools...@googlegroups.com.