Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 beta 3/9/83; site sdcrdcf.UUCP
Path: utzoo!linus!decvax!ittvax!dcdwest!sdcsvax!sdcrdcf!markb
From: markb@sdcrdcf.UUCP
Newsgroups: net.puzzle
Subject: Re: average wind direction puzzle
Message-ID: <1199@sdcrdcf.UUCP>
Date: Mon, 16-Jul-84 10:06:23 EDT
Article-I.D.: sdcrdcf.1199
Posted: Mon Jul 16 10:06:23 1984
Date-Received: Tue, 17-Jul-84 06:37:09 EDT
References: <146@ios.UUCP>
Reply-To: markb@sdcrdcf.UUCP (Mark Biggar)
Organization: System Development Corporation, Santa Monica
Lines: 27

In article <146@ios.UUCP> richa@ios.UUCP (Rich Altmaier) writes:
>Consider measuring wind direction with an instrument that gives compass
>angles (e.g. 0..359 degrees).  Suppose you wish to take several readings
>and give an average wind direction.  What's a good algorithm for this?
>
>If you take a set of readings and just numerically average the values,
>you may find you never get results near 0/360 and you get lobes near
>180 (which won't be correct).  Consider a reading set such as
>355, 359, 3, 1.  Average wind direction is about 359 degrees.
>
>	R. Altmaier
>	{qubix, oliveb, decwrl}!ios!richa

Numerical average only presents a problem if the range of measured values
spans across the 0-360 boundry.  Assuming that all the measurements are
within 180 degrees of each other, (otherwise wind direction is not very
well defined), the following procedure should do what you want.

1) if measurements do not span 0 degrees use numerical average.

1) if they span 0 degrees add 180 modulo 360 to each measurement
	take the numerical average and subtract 180 modulo 360 from
	the result

Mark Biggar
{allegra,burdvax,cbosgd,hplabs,ihnp4,sdcsvax}!sdcrdcf!markb