Stories
Slash Boxes
Comments

SoylentNews is people

posted by chromas on Thursday June 13 2019, @02:41PM   Printer-friendly
from the wx dept.

The long-awaited upgrade to the US weather forecast model is here

Weather forecasters need a ton of knowledge and a fair bit of experience with local weather patterns to do their job well. They also need a good forecast model. These computer models take in measurements from weather stations on the ground, satellites in orbit, and balloons in between and then simulate the physics of weather forward in time a few days.

For the first time in about 40 years, the guts of the US model got swapped out for something new today. The upgrade brings us a new "Finite-Volume Cubed-Sphere" (or FV3) dynamical core, which simulates the basic atmospheric physics at the heart of this endeavor, a change that has been in the works for a while.

The new core had its origins in simulating atmospheric chemistry but ended up being adapted into other models. A few years ago, it was selected to replace the old core in the US Global Forecast System model. And for more than a year now, the new version of the model has been running in parallel so its results could be compared to the operational model.

[...] The results have been a little mixed. The new core improves computational efficiency and allows some processes to be simulated at a higher resolution [...] But there have also been grumblings in the weather community over the past year about results that didn't seem so hot. For example, surface temperatures have been biased low in some situations, throwing off forecasts.


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 2, Informative) by Anonymous Coward on Thursday June 13 2019, @07:39PM (2 children)

    by Anonymous Coward on Thursday June 13 2019, @07:39PM (#855259)

    You are correct. The problem is that the world is not flat, so no matter how you try to model it in a flat plane, you end up squishing or stretching various areas (just look at the various projections with Tissot indicators). At smaller scales, that isn't a big problem because the errors introduced for short-term, local forecasts are not that large compared to those caused by chaotic influence from the observations outside the scope of your model (e.g. topography, lack of granularity in observations, influences of distant conditions, etc.).

    However, when you are trying to predict conditions for large areas or distant future ones, then the errors caused by using a plane or improper 3D atmospheric projection make a huge difference. Your distances and angles are all off, so the errors are very large. An example is Chicago; in order to properly model their weather a week in advance, you need to know what is going on in an area roughly stretching from the Arctic Circle to the Yucatan Peninsula and from just past the Western EEZ to 1/3 across the Atlantic. Even then, not all 3D grid systems are the same (the lat-lon system is hard to calculate, for example). So, the researchers did some studies on various trade-offs and decided on a particular "cubed-sphere" grid system, which represented the best balance, according to their goals.

    Starting Score:    0  points
    Moderation   +2  
       Informative=2, Total=2
    Extra 'Informative' Modifier   0  

    Total Score:   2  
  • (Score: 0) by Anonymous Coward on Thursday June 13 2019, @10:38PM (1 child)

    by Anonymous Coward on Thursday June 13 2019, @10:38PM (#855323)

    Bucky Fuller's icosahedral projection, Dymaxion Map, might be good enough for weather? Of course the separated edges need to be linked together computationally.

    One link is https://www.gislounge.com/dymaxion-map-projection/ [gislounge.com]

    • (Score: 0) by Anonymous Coward on Friday June 14 2019, @12:20AM

      by Anonymous Coward on Friday June 14 2019, @12:20AM (#855357)

      I just looked at that one, and I don't really see the benefit. For predicting a week in advance for any point in the CONUS, you need to look at 9 of the 20 regions. To use native atmospheric calculations requires 5 dimensions and lookup tables, and you still have distortions in the model larger than many common alternatives. Now, you could say, "just use lat-long." The problem with doing so is that you are no longer using the projection, but rather the geodetic model underlying that coordinate system. Plus, as stated, there are problems with lat-long computationally that make most cubed-sphere systems better.