[Ncep.list.idp.mrms.users] [awips2dev] MRMS
Michael Magsig - NOAA Federal
michael.a.magsig at noaa.gov
Fri Jun 22 19:38:55 UTC 2018
Hi Jeff,
Have you downloaded our AWIPS bundles from RAC to compare the precip
sources? Here is the RAC Procedures Download Page (note you can find
this on AWIPS by doing a manual AIR search for rac using the right click
on any CAVE text legend):
https://vlab.ncep.noaa.gov/web/wdtd/racproc
We have a methodology for quickly assessing differences in precip type
for choosing the best precip source and identifying things like melting
hail contamination that causes high biases in Dual Pol QPE. A difference
like this will stick out like a sore thumb in the instantaneous rate six
panel (see image below). You should be able to see how the rates compare
to Legacy and Dual Pol and see if there is anything unusual in the
patterns of surface precip type. Most likely it is triggering the
tropical convective if it is overestimating. Watch for temporal
continuity in SPT during the period of heaviest rainfall and see if it
is missing any hail signals in the base data.
I've been looking at a number of heavy rainfall events throughout the
country for our flash flood training development, and my perception is
that MRMS tends to perform a little better than Dual Pol which performs
a bit better than Legacy most of the time but it varies by
environment/season and can change during an event, so we recommend
starting each event assessing if you have any significant differences
between sources and choosing a source that looks to be the most
reasonable to start with. Then spot check your precip sources routinely
during an event to see who is doing best where it matters most. MRMS
typically shines in hail contamination events because its rate cap is
2"/hr for hail which usually lowers MRMS estimates significantly (which
is fine until you have heavy rainfall rates >2"/hr with hail). MRMS
overestimates like your case are usually tropical convective SPT when it
shouldn't be or MRMS not identifying hail.
Differences of 10-25% are common among all our operational precip
sources, and I routinely find differences of 50% or larger at times
during heavy rainfall. The biases can sometimes vary spatially and
temporally in an event (and during the seasons) for all precip sources,
so we preach using all of your observations and giving more weight to
those in the strongest cores for flash flood warnings, and only use the
mean field bias corrections or the MRMS spatially varying corrections as
a first guess. Also pay extra attention to MRMS biases because those
biases will propagate to the FLASH streamflows.
Accurate QPE is still a challenge for all of today's algorithms, and
those who rely too heavily on the precip source that worked in the
previous event are going to be susceptible to unpleasant surprises (I
see that a lot actually with many offices). Maybe we will have more
consistent separation for MRMS when it starts using Dual Pol variables
in one of the next versions, but I foresee the best warning performance
coming from carefully selecting and re-assessing your precip source and
blending that with the new streamflow products.
At least the MRMS overestimates out west should be addressed now that
11.5 has allowed continental ZRs instead of just tropical ZR for heavy
rainfall. Feel free to share some of your MRMS screen captures with me
offline, as I enjoy the challenge of quickly getting to the bottom line
for the differences in QPE.
Mike
On 6/22/2018 12:55 PM, Jeffrey Hovis - NOAA Federal wrote:
> All,
>
> I want to thank everyone for answering. It did answer some of my
> questions. I am going to read through everything and see where
> I go with any additional questions.
>
>
> Jeff
>
>
> On Fri, Jun 22, 2018 at 10:54 AM, Joseph Moore - NOAA Federal
> <joseph.moore at noaa.gov <mailto:joseph.moore at noaa.gov>> wrote:
>
> Jeffery, I'll definitely agree with you that for where the most
> precip fell, MRMS was on the hot side (mrms-radar-only.PNG). The
> gauge-adjusted product doesn't seem to have really changed much of
> that (mrms-gauge-adj.PNG), though looking at the diagnostic
> provided by MRMS (the /Gauge Influence Index/, viewable on MRMS Op
> Product Viewer site) there doesn't appear to be /any /gauge
> correction going on anywhere right now or in the past 24 hours for
> any precip accumulation timeframe. (*I think the issue of no gauge
> adjustment happening is an actual problem with MRMS, so I'm
> looping the MRMS Users email list on this thread*.)
>
> Compared to the RLX radar (STA, STA-hires, and STP images), the
> MRMS data certainly has a broader area of 3"+ and a fairly broad
> area of 4"+ compared to the legacy and dual-pol estimates. So even
> compared to other precip estimates, MRMS seems on the high side.
> Why the MRMS estimates are running too hot? As I said in my first
> message, it is probably a case of mis-classification of the type
> of rainfall occurring - maybe using the tropical algorithm when it
> should be using a standard convection one. Looking through the
> Reflectivity Cube (see the Op Product Viewer), during the most
> intense parts of the rain yesterday MRMS certainly seems to have
> ingested your radar and surrounding radars correctly with no gross
> errors obvious to me. The MRMS Ops folks on the Users listserv may
> be able to diagnose what's going on in this case better than I can.
>
> We can't expect MRMS to be perfect, just like we know the
> radar-based legacy and dual-pol estimates are going to have errors
> for various reasons. For what it's worth, on our recent heavy
> rainfall event last weekend MRMS absolutely /nailed /the precip
> estimation with no strong high or low bias - it got the maximum
> amounts just about as spot-on as you can expect... and that was
> with our primary radar down! (MPX's radar provided sufficient
> radar data to help fill in the gap to our radar being down, but at
> a higher elevation than DLH would have been scanning if it were up.)
>
> If there is a noticeable problem with MRMS data, this should be
> reported to the MRMS/NCEP folks so they can investigate (I think
> /ncep.list.idp.mrms.users at noaa.gov
> <mailto:ncep.list.idp.mrms.users at noaa.gov>/ is the best choice but
> someone else chime in there's a better option!)
>
> -Joe
>
> MRMS Op Product Viewer:
> https://mrms.nssl.noaa.gov/qvs/product_viewer.php
> <https://mrms.nssl.noaa.gov/qvs/product_viewer.php>
> WRH Wx and Hazard Viewer: https://www.wrh.noaa.gov/map/
>
>
>
> On Fri, Jun 22, 2018 at 9:19 AM, Timothy Humphrey - NOAA Federal
> <timothy.humphrey at noaa.gov <mailto:timothy.humphrey at noaa.gov>> wrote:
>
> Over the past several days here at the Lake Charles office, we
> also noticed that the MRMS-QPE Radar only was running quite
> hot compared to our individual radars' estimates and surface
> observations. Comparing this morning's 12Z 72 hour MRMS
> estimates to a variety of observation sites including
> ASOS/ALERT/COOP confirmed this with a few comparisons listed
> below:
>
> JYDT2: 7.91" MRMS: 10.33"
> JYET2: 7.76" MRMS: 13.18"
> JYGT2: 6.38" MRMS: 7.48"
> KBPT: 5.79" MRMS: 8.11"
> ORNT2: 4.91" MRMS: 6.06"
>
> Based on our experience, the QPE-Radar Only seemed that it was
> running much hotter compared to previous events and we were
> also curious what might have been causing such a large
> difference. Our concern is that these large differences could
> result in derived FLASH products being less reliable for
> warning decision making.
>
> On Fri, Jun 22, 2018 at 8:54 AM, Jeffrey Hovis - NOAA Federal
> <jeffrey.hovis at noaa.gov <mailto:jeffrey.hovis at noaa.gov>> wrote:
>
> Jack,
>
> It was approximately 10 miles from the area of heaviest
> rainfall to the radar site (RLX) . As you indicated,
> COOPS are not included in the MRMS data.
>
> I just wanted to see what might have caused the large
> difference in rainfall amounts.
>
> We issued a Flash Flood Warning based on the MRMS data.
> It was a good warning as we did have lots of flash
> flooding reports.
>
> Jeff
>
> On Fri, Jun 22, 2018 at 9:42 AM, Jack Settelmaier - NOAA
> Federal <jack.settelmaier at noaa.gov
> <mailto:jack.settelmaier at noaa.gov>> wrote:
>
> I may be mistaken, but I'm guessing MRMS does not use
> COOP data in its algorithms, as it's more real-time
> (not QCd too much) and mostly just uses radar data.
> How far from the nearest radar was the site?
>
> https://training.weather.gov/wdtd/courses/MRMS/index.php
> <https://training.weather.gov/wdtd/courses/MRMS/index.php>
>
>
> On Fri, Jun 22, 2018 at 8:10 AM, Jeffrey Hovis - NOAA
> Federal <jeffrey.hovis at noaa.gov
> <mailto:jeffrey.hovis at noaa.gov>> wrote:
>
> All,
>
> I just wanted to give you an update. When I came
> into the office this morning, I checked the MRMS
> 24 hour QPE amount against our COOPs, We had a
> COOP that was located in the area of heavy rain.
> MRMS indicated that rainfall amounts between 3.54
> and 5.11 inches near the location of our COOP. Our
> COOP actually reported 2.20 inches of rain.
> This is a very big difference.
>
> There was Flash Flooding in the area. However if
> the MRMS data had verified, the flooding would
> likely have been much worse.
>
> Jeff
>
>
> On Thu, Jun 21, 2018 at 3:23 PM, Jeffrey S. Hovis
> <jeffrey.hovis at noaa.gov
> <mailto:jeffrey.hovis at noaa.gov>> wrote:
>
> All,
>
> We are currently experiencing an area of heavy
> rain north of our office. The MRMS QPE-Radar
> Only product is indicating as much as 3.42
> inches of rain had fallen over a location in
> the past 3 hours. However, none of the
> surrounding radars are indicating that much
> rain has fallen. In fact, the highest 3 hour
> rainfall amount based on radar that I have
> found is closer to 2.5 inches.
>
> What could be causing this difference between
> these two products?
>
> Jeffrey Hovis
>
>
>
>
>
> --
> Jack Settelmaier
> Digital Techniques Meteorologist
> NOAA/NWS, Southern Region HQ
> Fort Worth, TX
> Work: 682 703 3685
>
>
>
>
>
> --
> Tim Humphrey
>
> Meteorologist
> National Weather Service
> Lake Charles, Louisiana
> 337.477.5285
>
> Follow us on Facebook
> <http://www.facebook.com/NWSLakeCharles>, Twitter
> <http://www.twitter.com/NWSLakeCharles> and Youtube
> <http://www.youtube.com/user/NWSLakeCharles>
>
>
>
>
> --
> *Joseph J. Moore*
> Meteorologist | WFO Duluth Social Media & IDSS Program Leader |
> Open Source GIS Evangelist
> NOAA/National Weather Service Duluth, MN
>
>
--
Michael A. Magsig - KD5YKJ
Meteorologist Instructor
Warning Decision Training Division
120 David L. Boren Blvd, Suite 2640, Norman, Oklahoma, 73072
Email: Michael.A.Magsig at noaa.gov Phone:(405)-325-2995 Fax:(405)-325-3203
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://www.lstsrv.ncep.noaa.gov/pipermail/ncep.list.idp.mrms.users/attachments/20180622/8afcb2fd/attachment-0001.html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dbdhoonnhicaiimg.png
Type: image/png
Size: 890587 bytes
Desc: not available
Url : https://www.lstsrv.ncep.noaa.gov/pipermail/ncep.list.idp.mrms.users/attachments/20180622/8afcb2fd/attachment-0001.png
More information about the Ncep.list.idp.mrms.users
mailing list