TIM Notes 28-29 Jun 04
Action items are presented in bold text.
Presentations open in another window, so you can read the notes while
reviewing the slides.
Present: Mark Wheeler, Lowell Schaffert, Charlotte Johnson, Chuck Fain,
Pete Conant, Stephen Cocks, Carol Adams, Susan Williams, Joe Wakefield,
Herb Grote, Phil McDonald.
What's new - Joe
FSL RSA progress since September meeting.
- ORPG going to Linux was a surprise to some
- Maj Cocks expressed concern at slow performance of 3-d lightning. Mark
noted that the Briefing Tool also has some performance issues.
- Pete asked what is meant by lightning processing on AWIPS. Lowell
explained that MARSS generates CSV-format bulletins which are processed
for display by AWIPS.
(Dave Sautter and Brian Bartram joined the meeting at this point.)
NWS AWIPS changes since OB1.
- Pete asked if the RSA merge to OB5 baseline means that future enhancements
will automatically get in, too. Joe responds no, but noted that it should
be routine in the future for NWS to agree to accept such additions.
- Steve wants WarnGen for local use (targeting specific locations on the
range), and also wants to include IFPS/GFE.
(Tim Wilfong and Phil Wildhagen arrived at this time.)
Steve said tabular text does not work on CCTV - it's fuzzy. He will
check to see if using large font addresses the problem.
Can we get both sides of the towers? Lowell explained that only the
windward side is sent to AWIPS - an algorithm in the ingest system decides
which one to send. Steve will revisit the issue of seeing both sets of
instruments. Lowell suggested a possible solution of adding a text
view of the data in the ingest system - not in AWIPS.
Levels listed in the tabular text are still wrong. Steve will provide
the actual levels. (We know 12 and 54; there's still a question about
some of the higher levels.)
If a subset of data are selected from the Scale pull-down, the label
should so indicate.
WR folks would like to do areas (North Base/South Base). FSL will do
so when provided with a list or map by WR.
Mark noted that FSL's sample data for tall towers are incorrect. Towers
2, 6, and 110 go only to 200'.
(Mike Schmeiser and Don Coleman came in during this discussion.)
ER will check performance of tabular text.
Chuck pointed out the importance of having the LWOs at these meetings.
We all tried...
FSL will post field mill tabular text examples on the Web for review
Archive server GUI - Susan
Lowell wanted to make it clear that this archive server is a stand-alone
system and is not related to the general RSA archive. It records only the
data from the AWIPS data disks.
Mark suggested adding a 9400 MB option to anticipate future hardware.
Users want to eliminate the option of deleting the tar files after
Steve asks if we need to remove files to make space.
Steve: Will selecting "archive" slow down my AWIPS? Lowell and
Susan answer No, as it's a separate server.
Mark suggests compressing the seven-day-old data automatically,
before overwriting it.
AX system needs to include a cleanup function to remove cases that
are not of interest. Also, the case limit still needs to be determined.
Mike (and others) wants a visual indication that we're viewing
archived data - e.g., with a label or color change in the frame title.
Mark asked about a ticking clock. Lowell notes that WES (NWS' playback
system) was specifically excluded from RSA, in favor of the replay on office
MARSS and other systems are on AS2.1 now. Only AWIPS (data server,
workstations, model cluster) remains on 7.3. Our task calls for converting
these in Mod 3 of the MOU.
Mark: Will Informix run on 2.1?
The compute nodes on the model cluster will not run AS2.1, so they
will have to be replaced.
Present: Chuck, Mark, Steve, Don, Mike, Phil W., Pete, Dave, Lowell,
Charlotte, Brian, Carol, Tim, Joe, John, Steve Albers.
(Depending on your browser, you may wish to adjust the width of the
window to view John & Steve's presentation.)
Steve C.: Isn't the 1 km domain too small? Has this been discussed?
John responded that the 1.1 km grid uses the 3.3 for boundary conditions,
which in turn uses the 10 km for its boundary. Lowell noted that, yes, it
was not only discussed, but was a requirement. Mike believes that the 1km
grid is more important on the west coast due to terrain. Steve hopes that
the domain size can be increased - perhaps this can happen with the model
cluster upgrade. Tim and Lowell point out that the hardware was speced to
meet the requirement. John suggests that running dry is an option that
would allow using a bigger domain.
John noted that FSL looks at the results every day - sometimes it's a
disaster, but usually the 1 km is good, especially for sea breeze. Once
convection starts, all bets are off - can't resolve gust fronts with
instrumentation. FSL is looking for feedback, once the users get their
hands on it routinely.
Steve: Will it be easy to configure the model once it's delivered?
Forecast length, resolution, grid location... John says, yes, this is all
in namelists - a couple of days of training will allow a local site
administrator to change these at will.
With improvements to NCEP models, it may be possible to eliminate the
10 km domain and expand the other two a bit.
Steve asks about radars - all FL radars? John says Yes, from SBN
(Level III data). Can accommodate Level II when they become available.
Steve A. discusses the local data sources. He notes that our RSA LAPS page shows the
cold temps near the Cape which is a result of using the synthetic data -
it illustrates that we are ingesting and using the data.
Lowell asks Pete about the status of the 50MHz profiler at ER. Vaisala
is to be on site 19 Jul to install a unit upgrade. Expect it to be fully
back on line for testing by the end of October. Interface to RSA should
be available in August.
Steve (A.) pointed out the double entry for lowest data level. Joe
will check the synthetic data for duplicate 15m report. [Found and
removed 7 Jul. JSW]
Steve will contact Mike Schmeiser re getting an account to dial in
for AMPS data.
Steve C. says when it rains, minisodar data get bad. Mike says they
see it, too, when it rains.
(Note that this is the same presentation as above. MM5 starts on Slide
Steve asks about the utility of SST climatology. Dave and Lowell say
Eta grids are due in a month, including SST.
John notes that Schultz II microphysics allows convection to start
with subsaturated grid boxes. This produces a much more gentle initiation
of convection tan most physics packages.
Verification of precip using Schultz µphysics shows significant
skill score improvements. John notes in particular that MM5+ outperforms
Eta in the 0-3 and 0-6h periods. At 0-12h, Eta nearly catches up - it just
doesn't spin up precipitation quickly.
Re the 41 sigma levels, Steve asks about instability with closely
packed near-sfc layers. John says one has to make sure the time step is
short enough to avoid having vertical motions cross layers. The current
30 seconds works.
Mark notes that running LAPS more often than once per hour has proven
effective at JAX and SMG.
John shows how long it takes to run the models. The 1 km is slow. Tim
wonders why we run out to 24h at all. There are many trade-offs to consider.
Latest version is 2.1, delivered to LM 7/03. Lowell says it has not
been installed at WR.
Mark asks if one of the old (replaced) compute nodes could be used to
run the verification? Lowell says a) don't expect them to be available and
b) AS2.1 won't run on 'em.
Steve needs to provide input to developers for what should be in
the verification package.
Steve will get the model-availability deficiencies to Jackie.
(Clay says some things aren't available on AWIPS, or not on the right
Briefing Tool - Herb
Herb said that he is in the process of testing a major revision of
the tool for Homeland Security work, but there are no plans at the moment
to put that in RSA.
Steve notes that Johnny and Clay have actually used the Tool to create
some briefings and have some feedback to provide.
Can LWOs develop templates? Herb says it can be done and he's done some
training at both Ranges. It's mostly HTML and is documented in the
Tool User's Guide, Section 4.1.2.
Mark: BT has reduced data availability vice AWIPS. Will we get the rest
of the models, e.g.? Herb says no, he does not intend to duplicate AWIPS.
Chuck again notes that we need to get LWOs and FSL folks together to
figure out what's needed for templates.
ORPG Build 6 - Darien
The 88D ROC is making Build 6 available on Linux, so we'll skip Builds
4 and 5 for RSA.
A concern is that the T1 between KMLB and the Cape ORPG may not support
the data needs of both AWIPS display and MM5, in particular with future
VCPs (faster volumes). Tihs is post Build 6.
The AMU needs radar data, too. How?
3-d Lightning - Phil
MarySue Schultz and Susan Williams joined the group for this session.
Phil provided a handout summarizing the changes since September, to wit:
The display retains 30 minutes of data, or on the order of 1M events.
At about 20 bytes per, the application needs about 20MB memory for its
data. Adding graphics, 3-d likes to have about 100MB available. Phil notes
that he concentrated more on ease of understanding the code rather than
trying to make it more memory efficient. He mentioned several ways that
the application could be made more responsive - use hardware rendering;
trim out parts of VIS5D that are not needed; reduce the redundant plotting
of events that are effectively collocated. (None of these is planned.)
Discussion focuses on what to do when zoomed in. Johnny had requested
something to indicate where other events are happening, possibly by piling
them on the perimeter or just tagging the perimeter. Phil's suggestion on
the handout of using hatching and a split histogram is
generally deemed acceptable.
- Current data show at the front/top, rather than geographically. (Data
are ordered by time in the paint process.)
- When one zooms/pans the top view, the S and W views sync up.
- The display now connects to a real-time data stream.
Mark: There are still times when the system gets slow, particularly
with the Briefing Tool. Discussion notes recent network reconfigurations
and problems that Lowell believes have been solved. Chuck suggests that
someone(s) from FSL needs to visit the Ranges to work directly with the
users on this issue. Lowell argues that we need a repeatable scenario.
Mark will solicit "slowdown scenarios" from LWOs.
Mark: Can we hook a one-head AMU machine to the operations network?
This would allow AMU user to view radar and local model data.
Joe: What about FSL's local data access? Lowell says the EDS is running,
but no data go there because any arriving data would be forwarded to the
AMU, which presently gets data from the FTP server. This was logged as a
PR in January (?) and is not on the to-be-resolved schedule yet.
Don reviewed some DRs. Lowell noted that OB1 will be installed at WR
soon. Problems without a ready solution (or already fixed in OB1 include:
Joe asked Lowell why AS2.1 is being used, when it's been superseded
by ES3.0. The answer is that the SPO directed LM to use 2.1.
Steve wants time series labels to stick with "round" numbers like :00
and :05 rather than go to 3 minutes or 11 minutes, etc.
- 217: LF03 ASOS sampling incorrect [may be fixed - this is the station
that lacks most instrumentation].
- 226: Composite reflectivity not available on State scale.
- 243: Load time series of three fields and try to capture the display
in the Briefing Tool - it fails. Apparently, this works OK for two fields.
Herb tried it once and it failed, but a second try succeeded. He thinks
it may be a timing issue with auto-update on the D2D.
Schedule and Tasks - Dave and
Mod 3 of the FSL-LM MOU includes tasks for ingest/display (Work Element
A) and analysis/model (Work Element B). From Element A:
As for Work Element B, John plans to deliver new versions of LAPS/MM5
in July 04 and January 05.
- Merge with AWIPS OB3. NWS will begin fielding OB4 in September. In light
of this, we combine this with Item 5.
- Stage ORPG Build 4. As noted earlier, ORPG Build 6 will run on Linux.
This task is combined with Item 7.
- Stage RedHat Enterprise AS2.1. FSL plans to complete this on their
development systems by the end of September. LM expects to have the new
compute nodes for the model available by mid August.
Related to this topic, there was an extended discussion of sustainment.
Tim pointed out that end users need to advocate (with the SLRSC SPO in
Colorado Springs) that RSA be kept up in sync with NWS. The three-year
hardware refreshment plan that NWS is on saves money by avoiding maintenance
costs. In addition, keeping in sync allows RSA to leverage all of the work
done by NWS in developing AWIPS and ORPG.
- Field Mill tabular display. Work will start mid July, with completion
in December. An intermediate version lacking the timer will probably be
included in the initial OB4 delivery.
- Merge with AWIPS OB4. This work will start in July and be completed by
the end of September.
- Briefing Tool enhancements. Herb expects to complete the tasks on his
list by the end of December.
- Stage ORPG Build 5. FSL could stage ORPG Build 6 by mid August, but the
RDA won't support Build 6 until September, says Darien. This will be
delivered with OB4.