[Geysers] Giant?

[Ralph Taylor]

I also agree that some form of check is needed.  If the data is
unreliable, or more importantly thought to be unreliable, its value is
diminished.

The paper logbook in the OFVC (now the OFVEC) has had similar problems
over the years.  The online transcribed logs have some annotations,
sometimes just a question mark, when data that appears suspicious or
inconsistent appeared.

Lynn Stephens and I looked at several geysers for which there were
both logbook entries and electronic data logger records and found
numerous cases where the wrong geyser was reported or when an eruption
was entered on the wrong day.

Some of these errors were artifacts of the physical setup (the
reporter may have been interrupted by someone else needing the logbook
and then resumes entering data without noticing that the page was
turned.

The online logs can and probably do (I haven't personally used them or
looked at them) prevent many of these errors, but it is likely that
some other kinds of errors can creep in.  However, there should be a
way to enter data that is questioned by automatic checks because, as
we well know, geysers can do whatever they want!

Probably the best data is a synthesis of the OFVEC log, the webcam
logs, the online logs, and the electronic record (if it is available).
 I have often used the webcam log to help decide whether some event in
the temperature log is actually an eruption.

Data consistency or reasonableness checks can be automated, and can
help by prompting the person entering data.  Udo Freund made some good
points about this also.  Flagging reports and developing a metric for
posters' reliability seems a good idea to me.

On 10/12/11, David Schwarz <david.schwarz at alumni.duke.edu> wrote:
>    I agree that, elitist as it sounds, estimating the reliability of reports
> at least partially based on who reported them makes a lot of sense.
>
>    For a lot of the most frequently reported geysers, it would also be
> possible to do basic smoke checks for whether a time is reasonable, and
> automatically flag questionable reports for review.  For example, two
> eruptions of Plume ten minutes apart, or a four-hour Riverside interval, or
> even a nine-hour double Riverside interval would be suspicious.  So would
> Penta during or shortly after Sawmill (almost certainly a steam phase
> eruption).  Infrequent geysers like Splendid, Morning, Giant or even North
> Goggles in recent years could be flagged automatically.
>
>    Such analysis could also catch sudden changes in activity, like when Lion
> went overnight from have series of three eruptions or fewer to as many as 30
> a couple of years ago.
>
> David Schwarz
>