10 January 2002:  Time conventions on models:

From Tony Bannister

Paul,
I understand the analysis part but I am still stuck on the prognosis times.

I am currently looking at the mesolaps which has a windchange west of Melbourne
at 0z tomorrow, presently do I take this as the meso thinks it is 11am (AEDT) or
10am or 9am?  Talking to my NSW collegues they are punting for 9am, I would like it
to be 11am (to make our forecast for Melbourne temp look half reasonable).  Whatever
the time is, is this the same for GASP and LAPS?

Do you then know how EC, US,UK prognosis times fit into the picture?

Yours in anticipation,

Paul Stewart wrote:

> RE: BASE TIMES IN NMOC'S OPERATIONAL NWP SYSTEMS
>
> Dear Colleagues,
>
> The following is an attempt to answer Bob Moore's query as to "What is the  real
> time on the models?"
>
> The times for the various operational NWP systems in NMOC are still based on 11
> and 23 UTC (and also 05 and 17UTC).
>
> However, this does not mean that only observational data for these specific
> times are used. The current operational assimilation schemes (LAPS, TLAPS,
> including TC_LAPS, and GASP) use 6-hourly data insertions. This effectively
> means that 6 hours of observational data is used in generating an analysis at a
> given time (ie at 05, 11, 17 or 23 UTC).
>
> Currently for a given analysis time, the time window used is from -2 to +4
> hours. Thus, for example, observations from 09 UTC to 15 UTC are used in the
> analysis for 11 UTC. (It is further noted that the observations, within the 6
> hour window, are effectively weighted according to their time difference from
> the nominal analysis time.)
>
> Over the years, a number of changes have been implemented with the aim of
> eventually using the base times: 00, 06, 12 and 18 UTC consistently  throughout
> the operational system in NMOC. These have included:
>
> . the adjustment of chart labels so that base times for analyses were rounded
> off to 00, 06, 12 and 18 UTC and valid forecast times adjusted accordingly;
> . the use of 00, 06, 12 and 18 UTC in the actual data set names within the real
> time data base (which were introduced for GASP several years ago and have,  just
> in the last few months, been used for the various LAPS systems).
>
> However, the observational extraction procedure and the internal files
> (including those in NetCDF form), used within the operational NWP systems,
still have base times of 05, 11, 17 and 23 UTC.
>
> It is only when these latter procedures and file structures are changed that we
> can say that the base times are 00, 06, 12 and 18 UTC. (It is planned that this
> will happen during 2002.)

From Robin Bowen

hi Tony et al

> However, the observational extraction procedure and the internal files
> (including those in NetCDF form), used within the operational NWP systems, still
> have base times of 05, 11, 17 and 23 UTC.

so where your chart says 00Z the actual time is 23Z, ie one hour prior !
and the AEDT time is 10am  ( if not daylight saving then it would be 9am )
this is the same for LAPS and GASP.

I will leave Paul or someone else to answer concerning EC, US,UK prognosis
times !

cheers

Robin

From Phil Davill

Hi all,
 Surely if the model time is for 23Z it should be labelled for 23Z not one
hour later. If we want 00Z why not output the 00Z data (not rename the
23Z to 00Z) and run the model out to 49 hours instead of 48 hours.
 What we are currently doing sounds very unscientific. It would be
interesting to see what a smart lawyer would do to us in a court
case.
      Phil Davill

From Tony Bannister
Thanks Robin,
So is the same for all the hourly progs, eg is 07z really 5pm AEDT?  Phil Davil
raises a good point, also if you took a straw poll amongst forecasters I bet they took
the time to be correct (ie 07z being 6pm AEDT).  With the models getting better at
timing of windchanges we look carefully at what the model is saying, and at what time
it is saying it.  I was using hourly meso5k heavily in Sydney last week forecasting
for the NSW bushfires. The prog times should indicate the correct time.

Graham Mills

Regrettably, I'm forced to comment -

As described by others, the base time for the LAPS, meso etc
models is 23Z (or 11Z etc). Thus only data at that time gets full
weight, and obs 1 or 2 hours off get less weight. However, its more
complex than that - the analysis is a correction of a background
field, which is the short-term (6-hour) model forecast, and for a 23Z
analyis is valid at 23Z (this means 23Z here). If the data is at 22Z
(eg the synops) then the increment is the difference between a 23Z
guess and a 22Z obs, downweighted by the time de-correlation so
that the amplitude of any correction is down-weighted. Its pretty
easy to see than a progressive trough will also introduce a phase
error here too, with subsequent unknown consequences. Thus, as
I've said at training sessions many times in the past, the numerical
analysis is produced as an initial state for the forecast model, not as
a substitute for a meticulous manual analysis.

Also critical for the meso-scale models is the fact that the model's
radiation, and thus its diurnal heating cycle, is tied to the 23Z base
time. This can be critical in situations like today's forecast for
Victoria, as the rate of movement and evolving structures of these
changes is strongly modified by the diurnal cycle of land-sea
heating contrast. Sea-breeze circulation timings are also obviously
affected, as would be even tempereature forecasts for a given time.

Thus a forecast saying 00Z on NMC's web site, and I think on the
MCIDAS files if I'm interpreting the plots on VRO's web site
correctly, really means 23Z. Just to be really confusing, if you are
looking at the pretty coloured 5km meso-LAPS charts that come
through BMRC's web site, then the times are not incremented by
1hour, so a 23Z-valid forecast there is the same time as the 00Z-
valid (labelled) chart on NMOc's web site, and really means 23Z.

Confusing eh? Its hard enough out there without having to do this
sort of mental arithmetic, as well as the extra calculations for daylight
savings time, so be careful.

As a challenge, who'd care to explain all this to the media, or to a
coroner if you had to?

Graham (not really much of a guru) Mills
 
 

David Thomas

Dear all,
Terry Hart is away at the moment so I will make the following comment on his
behalf. He may put out a more current and informed view on his return on 15
January. Paul Stewart is also on leave.

Firstly, let me clarify; all the models we receive from outside of Australia
are, and have been for as long as I know, based on 00 & 12z. Some of the
newer models also run on 06 adn 18z. International observing times are also
built around the 00 & 12z. Australia really has been the odd one out in the
global modelling community.

NMOC has been driving the push to move from the traditional use of 23/11z in
Australian models to 00/12z to be inline with the rest of the world. This
has been a mammoth project and an incremental one. Some of the steps have
been taken based on opportunity (new RTDS2) , others based on need. Paul
Stewart's email earlier describes the plan quite well, and you can see from
Robin and Graham's comments, just how complex this is. However, most things
are now in place to take the final step to be fully converted to 00/12z
early this year.

regards

David Thomas
Manager Central Operations (SRCO) & a/STAN