Ticker for May 9, 2006

                
MESONET TICKER ... MESONET TICKER ... MESONET TICKER ... MESONET TICKER ...
May 9, 2006        May 9, 2006        May 9, 2006        May 9, 2006


Some Quality Time with the Ticker

Whether we recognize it or not, all of us involved in selective
processes balance two types of errors: Type I (false positives -
mistakenly identifying "no" as "yes") versus Type II (false
negatives - mistakenly identifying "yes" as "no"). Gardeners do it
when they try to pull as many weeds as they can without pulling
desirable plants. Managers try to hire as many good employees as
they can while minimizing the rate of bad ones.

Heck, if you read the Ticker by e-mail, your spam filter is a
perfect example: it throws out as many worthless e-mails as
possible, while preserving the meaningful ones. If, as a result of
this Ticker, you move your "annoying science blogs" setting to be
more restrictive, you have in essence told your spam filter to
shift the balance of power toward Type II ("I don't care how many
other e-mails you clumsily destroy, just rid me of this vile
Ticker!").

Well, ensuring quality data at the Oklahoma Mesonet (or any weather
network) involves an endless tweaking of this tug-of-war between
the error types. We are proud of our Quality Assurance (QA) system
here, because it constantly adapts and learns from its past
performance. But more importantly, it involves continuous oversight
from our QA meteorologists, who take the attitude that all data
are innocent until proven guilty. In other words, even if the
sophisticated QA algorithms identify an observation as "bad", that
datum gets to plead its case to the QA meteorologist. Most of the
time, the observations are indeed faulty. But sometimes, they are
very real, and the result of something really interesting.

And that's where the beauty of the system comes in.

So, let's imagine ourselves imagining things.

Now, imagine a gardener going through her pile of pulled weeds, just
to make sure none of them are actually desirable flowers. Now imagine
that gardener found a pretty pansy in the pile. Now imagine the
gardener putting the pansy back in its rightful place in the garden.

Now imagine that pansy being a 105 mph wind gust.

And that's what happened Friday. The Idabel Mesonet station reported
a wind gust of 105 mph Thursday afternoon, and the automated QA system
initially flagged it as "questionable". The data point stuck out so
much, compared to its neighbors and compared to Idabel's wind history,
that the automated QA didn't trust it. However, the Mesonet QA folks
gave the datum a fair trial and discovered that it was indeed a true
observation, and restored its rightful place in our archives.

(by the way, the gust ranks among the top five in the network's history)