Received: by zonker.hpl.hp.com (1.37.109.8/15.5+ECS 3.3+HPL1.1) id AA17223; Wed, 28 Sep 1994 14:25:54 -0700 Date: Wed, 28 Sep 1994 14:25:54 -0700 From: Ken Poulton <poulton@zonker.hpl.hp.com-DeleteThis> Return-Path: <poulton@zonker.hpl.hp.com-DeleteThis> Message-Id: <9409282125.AA17223@zonker.hpl.hp.com-DeleteThis> To: wind_talk@zonker.hpl.hp.com-DeleteThis Subject: Re: Windsight forecasts stats
> Unlike NOAA Windsights two meteorologists are making forecast specific
> forecasts for17 Bay Area sailing sites. When NOAA forecasts 25 knots for the
> Bay Area and some sites, say 3rd Ave., are dead while others like Berkeley
> are cranking they could say their forecast was "accurate" overall.
Note that I'm not talking about the Bay Area region forecasts; clearly
these cover too broad a region to be useful. I'm talking about the SFO
Terminal Forecast (FT product) and comparing it to SFO hourly readings.
This is forecasting in blocks of several hours for a single site with a
wind sensor on site, so it is quite comparable to what you are doing.
It also just happens to be the best sensor site for windsurfing on the
Peninsula.
Their correlation is lousy. I think you should be able to do better,
but I'd sure like to see the numbers. I have 2.5 years of forecast
and hourly data stored if you're interested.
What do you think is the best statistical measure of forecast correctness?
The std dev of forecast vs the average of the 3:00, 4:00 and 5:00 readings
is 4.7 knots!
Ken Poulton
poulton@opus.hpl.hp.com-DeleteThis
"I bet the human brain is a kludge." -- Marvin Minsky
This archive was generated by hypermail 2b29 : Mon Dec 10 2001 - 02:28:12 PST