Comments on: Calculating Wind Direction

By: An Interview with Dick Kusleika » Bacon Bits:

An Interview with Dick Kusleika » Bacon Bits: — Wed, 07 Jul 2010 06:09:00 +0000

[...] Dick frequently chooses topics that are arcane, but curiously appealing topics such as “Calculating Wind Direction” and “Charting [...]

By: Rob van Gelder

Rob van Gelder — Mon, 04 May 2009 02:04:00 +0000

I thought this looked familiar: http://www.dailydoseofexcel.com/archives/2005/08/15/wind-direction-chart/
Amazing to think.. this blog has been running for 4 years.

By: Jon Peltier

Jon Peltier — Thu, 30 Apr 2009 04:47:00 +0000

I wrote up a more graphical approach today, which still uses Dick's clever formula, but uses real charts instead of those silly pies to help describe his 180° double secret rotation trick. It also showed the error of T's ways. The post is called Calculating Wind Direction per DDoE.

By: Doug Jenkins

Doug Jenkins — Wed, 29 Apr 2009 10:36:00 +0000

The trouble with the Forecast() function is it fits a straight line to your data, which isn't necessarily what you want to do. For instance in your fourth example, if you enter something close to 5000 you will get a result of about 185, rather than 170, which I'm guessing is what you'd want. What you often need is a piecewise straight line interpolation; i.e. for values between 2000 and 5000 fit a straight line between those two values, and for between 5000 and 10000 use a different straight line. There should be an Excel function to do that, but there isn't. Plenty of UDFs that will do the job though (including at http://newtonexcelbach.wordpress.com/2008/08/10/intersections-interpolations-and-rotations/)

By: Jon Peltier

Jon Peltier — Wed, 29 Apr 2009 03:22:00 +0000

Dick – Pretty clever. My approach was adding 360° to “selected” directions, getting the FORECAST, and MODding by 360. Basically the same, except I eyeballed the “selected” directions, rather than coming up with an algorithm like your test that they are within 180°. By the time I got there, I decided your approach was as valid as mine.

T – I tried your way, and got irregular results. Converting to [-180,180] changes the directions in Dick’s fourth case from {190,170,200} to {-170,170,-160}. Instead of varying by 30°, they now vary by 340°. The R-squared value of the resulting FORECAST is 0.014, compared to 0.980 for the lowest R-squared of the other cases.

By: T

Tue, 28 Apr 2009 14:56:00 +0000

I converted the wind directions to range from [-180,180] by using atan2(cos(theta),sin(theta)). Then a linear interpolation like Forecast works just fine.