search

firelab / windninja

A diagnostic wind model developed for use in wildland fire modeling.
https://weather.firelab.org/windninja/
Other
115 stars 43 forks source link

Investigate National Blend Model #222

Open ksshannon opened 7 years ago

ksshannon commented 7 years ago

Possible new wx model.

TIN: http://www.nws.noaa.gov/os/notification/tin16-33nbmv2aac.htm

nwagenbrenner commented 7 years ago

Is this similar to/a replacement for NDFD? They sound very similar.

ksshannon commented 7 years ago

I think so, or that's at least how I perceived it. No CONUS yet. I'll put more notes on the ticket when it's available.

On Wed, Oct 19, 2016 at 2:21 PM, Natalie Wagenbrenner < notifications@github.com> wrote:

Is this similar to/a replacement for NDFD? They sound very similar.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/firelab/windninja/issues/222#issuecomment-254929203, or mute the thread https://github.com/notifications/unsubscribe-auth/AAwVDUFC-3Cpl4A0bNZya3OS3scKCymLks5q1nvggaJpZM4KbT3V .

Kyle

rnleach commented 7 years ago

It's not a replacement for NDFD, NDFD will still be there and will still be the official forecast. This is like another model, but it is a blend of several models. Notably the ECMWF will not be used in the blend the last I heard. This is for reasons related to the limitations of the contract NOAA has to license their use of ECMWF guidance.

It is produced on the same grid as the NDFD with an expanded area. It has the same 2.5km resolution, and does (or soon will) have the exact same topography behind it that the NDFD does. So the two should be directly comparable in a very straightforward way.

nwagenbrenner commented 7 years ago

Thanks for the clarification Ryan. So is the blend strictly based on models (no human intervention/adjustments like is done for NDFD)? Is the blend at the national level or is it done by the regions like NDFD and then "stitched" together?

rnleach commented 7 years ago

The blend is strictly models, and is done at a national center in Maryland for the whole CONUS. There is no forecaster input. However, it is bias corrected point-by-point using the RTMA as the analysis.

So for each point individually, each model has a bias correction applied based on recent verification against RTMA. Then, each model is given a weight also based on recent verification, and the bias corrected models are all blended via a weighted average. This is done point-by-point, so every 2.5KM grid cell has it's own unique bias correction and blend.

This is all still very new, and I'm not sure if they've made a final decision on the bias correction scheme or the scheme they use to assign weights based on verification. They are experimenting still. Also, the RTMA is still actively being developed and is not yet stable for some elements. It has problems, but there is a lot of effort going into it now, and it is getting better. Temperature, dew point, and humidity are probably in good shape, especially from a wildfire perspective (dry, hot, well mixed) for both the RTMA and Blend. Winds and wind gusts are currently a very active area of development, especially in the RTMA.

A goal of the project is eventually every NWS office will use this as the initial forecast, so they all start from the same point. Then, when the situation merits, individual offices will deviate from that standard, first guess forecast. That way everybody has a common starting point. Hopefully this will address the majority of the problems with "stitching" independent forecasts together. So eventually this will have a big impact on the NDFD, and in many areas will be exactly the same. But there's still a long way to go until that happens.

ksshannon commented 5 years ago

https://www.weather.gov/media/notification/pdfs/scn18-78nbmaab.pdf

rnleach commented 5 years ago

This is an ongoing project in NWS and has done well on temperatures and humidity. For winds it tends to have a low bias when compared to ASOS, but I'm not sure when compared to RAWS. RAWS use a 10 minute average, and so usually have a lower predominate wind speed when compared to ASOS, which use something like a 2 minute average. So it might work pretty well.

On Wed, Sep 19, 2018 at 3:29 PM Kyle Shannon notifications@github.com wrote:

https://www.weather.gov/media/notification/pdfs/scn18-78nbmaab.pdf

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/firelab/windninja/issues/222#issuecomment-422965156, or mute the thread https://github.com/notifications/unsubscribe-auth/AJSF1h90ueZn8uEiHF2Zv8mvDBRolwG3ks5ucrdJgaJpZM4KbT3V .