Paul, (etc)

Sorry for the delayed response.  I was out of town during the event.

We have been attempting to incorporate the 24 hour grid into the 6 hour files.  However, the 6-hour grids failed to be created properly for this hydrologic day and did not include it.  I do not know why, as any logs associated with this day have been purged.   We had been having some issues getting our grids to send properly since our AWIPS 17.2.1 update, but this looks like a different problem.  They have been created properly since that date.

So this explains why the URMA (which uses 6-hour grids) was different than the AHPS precipitation grid, and why it was lacking.

Do address this issue more...

We have been working on a technique to distribute that 24 hour total into the 6 hour grids.  This has been working pretty well, except for that date.  We are actually creating a distributed one hour grid which we supply to some of our external partners and which we can also use in our river model.   The technique uses GFE to distribute the 24 hour grid based on the timing of a variety of grids:  qpe, mrms, hrrr and ruc.  It then creats the 6 and 1 hour grids.  It works pretty well and shows a lot of promise, but is still under development.

I hope in the future you will see further improvements.  It wont be perfect, but we are getting as close as the science and office workload can support.

John

mbrfc

On Tue, Feb 20, 2018 at 8:30 PM, Ying Lin <ying.lin@noaa.gov> wrote:

Hi Paul,

[Thanks for emailing qpe.rtma.urma.feedback.vlab@noaa.gov, though VLab is down today]

AHPS at water.weather.gov/precip (24h only) and URMA (6h,1h) both come from the NCEP Stage IV.  The 24h (12z-12z) Stage IV mosaic uses 24h QPEs from MBRFC when available.  In the 24h ending at 12Z 15 Feb, it appears that the MBRFC 24h QPE is much more realistic than the 6h QPEs, perhaps benefitting from 24h gauges; it's also more consistent with the CBRFC 24h total (summed from 6h QPEs; CBRFC does not produce a separate 24h QPE).

Attached: 1) 24h QPE from MBRFC;  24h sum from 6h QPEs, from 2) MBRFC  and 3) CBRFC, for the 24h ending 12Z 15 Feb. 

Question for MBRFC: in the case when a 24h QPE is different - much improved - from the sum of the 6h QPEs, can the corresponding 6h QPEs be revised so that the correction made in the 24h QPE would be reflected in the 6h QPEs?

Ying

On 02/20/2018 06:07 PM, Paul Schlatter - NOAA Federal wrote:

All,

Resurrecting this string of emails in the hopes of narrowing the list of potential sources of the discontinuities in QPE.  The state of Colorado once again saw some pretty drastic differences across RFC boundaries on 2/15.  For the images in this email I will focus on the Colorado RFC and Missouri RFC boundary along the Park Range in northern Colorado.  It's the western border of Jackson County, eastern border of Routt County, bordered on the north by Wyoming.  On Feb 15th a nice snow event hit northern Colorado.  First, the AHPS 24 hr QPE for the 15th:

Next, the AHPS QPE for the 16th.  Note that most of the QPE fell from 00Z on the 15th through 00Z on the 16th, but AHPS doesn't allow 24hr amounts from 00Z to 00Z, so you'll have to mentally add the 2 images.

The Park Range is the N-S oriented high QPE that starts at the Wyoming border in the top center of both images.  Notice that pretty good QPE extends east of the max.  For west upslope events, the max in snowfall occurs along the terrain gradient on the west side up to the ridge tops, but a lot of QPE "blows over" to the east, sometimes all the way to the valley floor in central Jackson County.  Granted, it's FAR less in places like Walden but on the east slopes of the Park Range, high QPE can still occur.  In fact, a CoCoRAHS observer on the east side of Jackson County (well east of the traditional max along the ridge top) had 0.60 liquid from 12Z 2-15 to 12Z 2-16.  All that said, these AHPS graphics seem to capture the real QPE patterns and amounts well for northern Colorado.  Keep this in mind when you look at the URMA graphics next.

First, what we get in GFE.  We have 6 hour grids for URMA QPE.  This is a loop of the last 6 days of QPE.  You'll notice that at 06Z on 2-15 there is high QPE Park Range for the Colorado RFC side of the mountains but nothing on the Missouri RFC side.  This pattern of higher QPE in Colorado RFC vs. very little for Missouri RFC continues until 18Z on the 15th.  The east slopes of Jackson into the north park area of Jackson county never see any QPE for this event in URMA, despite a good amount shown in AHPS.  Again we feel AHPS is accurate while the URMA QPE for this event is not.

Finally, from the Veritas website the 24 QPE ending at 06Z 2-16 is shown below.

​I chose 06-06Z because that was the bulk of the precip event in Jackson County.  00Z to 00Z would have shown pretty much the same thing here.  What it shows again matches what we have for URMA in GFE.  It is in stark contrast to AHPS QPE.  It's clear this URMA image is not taking advantage of CoCoRAHS and the 0.60 amount 5-10 miles east of the max QPE.  Further east, across the Medicine Bow Mountains (eastern Jackson County/Western Larimer County west of Ft Collins) where AHPS has 0.3-0.6 total across the 2 days, URMA has 0.05 to 0.1.  I am not saying which is more correct in the Medicine Bows because there are very few obs there, and no CoCoRAHS obs at all.

The question remains:  Is URMA using the best available QPE from the Missouri Basin RFC?  Seems as though the AHPS page QPE has higher quality data for our mountain sites across northern Colorado.  On the Colorado RFC side of our area the URMA amounts look more realistic and in line with observations.

Thanks,

Paul

​
​
​

On Fri, Jan 12, 2018 at 7:13 PM, Ying Lin <ying.lin@noaa.gov> wrote:

Paul (S.), Dave, John and Paul (W.):

   Looking at the 6h Stage IV/URMA ending at 06Z 7th, and 24h ending at 12Z 7th generated here, mosaic last updated on 12:37Z on the 12th (see P1 of attached PDF - they look somewhat different from those from Veritas - not sure that it can all be explained by different color scales).  I think the outline in the precip fields didn't so much follow the RFC boundaries as followed the rain/no rain (or at least < 0.1mm) 24h gauge reports (P2); the odd "zero precip" area showing the contour of the northern tip of WGRFC does have nearly all zero (or at least < 0.1mm) 24h gauge reports. 

   Regardless of the validity of discontinuities in this particular case, there is no question that RFCs have different biases that show up in long term (e.g. seasonal) accumulations and there might not be an easy solution to that, even with the on-going efforts (AWIPS DCS #17911, MDC 16-032 etc.) - they might minimize day-to-day visible discontinuities, but differences in long term biases are harder to eliminate.

   The ConUS Stage IV/URMA is the RFC QPEs stitched together - a point on the ConUS domain is from the RFC domain it belongs to, no massaging/smoothing. 

   Currently the mosaicked RFC QPEs is still better overall than the MRMS (http://journals.ametsoc.org/doi/abs/10.1175/BAMS-D-14-00174.1), though of course if MRMS is used for URMA, there will not be RFC boundary discontinuities within the analysis domain (e.g. ConUS).  Apart from accuracy, there are also production reliability issues: every now and then there is a missing (gauge QC'd) MRMS hourly, and I've been told that at present the MRMS production cannot go back to re-do a missing/problem hour.  The MountainMapper MRMS we get here is still the "para" version and often problematic - the 6 one-hourly files covering 00-06Z 7 Jan are showing completely zero values.  I think in the future MRMS (perhaps Q3MS or versions beyond that) have the potential of providing the best whole-ConUS URMA, but it's not quite there yet.

Couple of questions:
1) How did you get your MRMS plots?  I haven't found MRMS on Veritas
2) Anyone Python users know how to add RFC boundaries (say using shapefile http://www.nws.noaa.gov/geodata/catalog/hydro/html/rfcbounds.htm) in Python plots?  Is there already some basemap 'drawrfcboundaries' module somewhere similar to 'drawcoastlines/drawstates'?   I haven't been able to draw the shapefile boundary, and ended up overlaying precip plots on top of the "RFC domain mask" plot, using a cellphone flashlight below to make out the boundaries (no light table).  There has got to be an easier way. 

Ying


On 01/09/2018 11:29 AM, Paul Schlatter - NOAA Federal wrote:

Hi John,

I think looking at CAVE data and EDEX is likely a red herring.  The point Dave was making is that the Veritas website, which displays the national URMA data in the lower left, shows the exact same discontinuities as the previous screen shots of RFC QPE he sent you that originated from our local GFE at BOU.  Via this evidence, he is ruling out any local config or display issues.  The outstanding issue is that there are there unrealistic discontinuities in the 6 hour RFC QPE at the RFC boundaries, and those same discontinuities are not showing up as much in the day 1 Stage IV data (perhaps because it is smoothed but we don't know for sure).  We do not use 6 hour RFC QPE as verification; that is not why we are sending along this feedback.  

The reason we care so much is that this 6 hour RFC QPE with unrealistic discontinuities gets into URMA, which in turn impacts the quality of the NBM.  It looks like MBRFC 6 hour QPE is computed differently than CBRFC and ABRFC in our area, which in turn impacts the NBM.  Is there a better QPE source to use as input to URMA?  Is there a systematic problem with how one of the RFC's computes QPE?  Or how the QPE is stitched together?  These are the types of questions we've been pondering.

To help illustrate the point about discontinuities , the MRMS Mountain Mapper 6 hour QPE, 6 hour, for the same time period 1/7/18 at 06 and 12Z, is below:

 

The MRMS mountain mapper does not show discontinuities along RFC boundaries.  Sure seems like for a 6 hour QPE this product would be nice to use in URMA?  I say that without knowing the constraints of URMA and whether the MRMS mountain mapper can meet that need or not, and I am not up to speed on the biases of the mountain mapper.  Main point is that there are operational alternatives out there that do not have discontinuities at the RFC boundaries that URMA could use if there is no way to solve the issue in the 6 hour RFC QPE.

Thanks,

Paul

​

-- 

Paul Schlatter
Science and Operations Officer
National Weather Service 

Denver/Boulder Weather Forecast Office

303-494-3210 x766

Paul.T.Schlatter@noaa.gov

On Tue, Jan 9, 2018 at 7:00 AM, John Lague - NOAA Federal <john.lague@noaa.gov> wrote:

Dave,

So the 6-hour grids have not changed?   Perhaps something in your localization is preveting this; QPE data is supposed to update when a new file comes in.

There are problems with the way QPE is decoded and stored by EDEX.  It isn't quite right.  Some of them never get stored right and end up as "GribModelxxxx".  Some data are just filed wrong.

Do you know how to use the Cave/d2d product browser?  Do you think you can find the actual grib file name under which they are stored?

Cave menu->Data Browser ->Product Browser,

Select Grib, Drill down to the various "QPE grids.  It might be storing as "QPE-RFC-KRF" or "QPE-XNAV-KRF". 

How are you displaying it from the Cave Menu.  When I select "6hr RFC Local Mosaic", what I get is actually one-hourly data.

John

On Tue, Jan 9, 2018 at 1:36 AM, David Barjenbruch - NOAA Federal <david.barjenbruch@noaa.gov> wrote:

John and Ying,

These are 6 hour QPE grids that I shared, and those particular grids ending at 06Z and 12Z in this case are where the trouble lies.  The 24 hour QPE at the AHPS site John referenced usually has less discontinuities (with the exception of the CBRFC and WGRFC Paul noted), so that's the difference I'm hoping we can track down as that may be leading to NationalBlend data quality issues.  

I checked again tonight, and the data remains unchanged.  The NationalBlend Veritas site also has the same data, so it's being distributed nationally and it's not just a local issue. I've attached the same verification data images from there for the same time periods, ending at 06Z and 12Z respectively on January 7, 2018.  The URMA QPE is located in the bottom left corner of each image. 

Thanks,

Dave

On Mon, Jan 8, 2018 at 9:06 AM, Paul Wolyn - NOAA Federal <paul.wolyn@noaa.gov> wrote:

I first noticed the discrepancies between RFC QPEs a few years ago, and I let others know about the QPE problems.  Attached is the QPE for Colorado for the last 180 days, ending this morning.  The ABRFC, MBRFC and CBRFC appear the match fairly well.   (Given the coarseness of the scale, it is difficult to see if there are still some differences at the borders.)  There still is a noticeable difference between the CBRFC and WGRFC in the San Juan Mountains.  

Thanks,

Paul

On Mon, Jan 8, 2018 at 6:57 AM, John Lague - NOAA Federal <john.lague@noaa.gov> wrote:

Your data not updating?  Or perhaps you ran this too early?  The AHPS precip looks much better:

The only data for mountainous areas comes from SNOTEL sites, which arrive later in the day and are difficult and time consuming to QC.  Also, there are very good reasons there may be discontinuities across a continental divide.

Also, I will probably so a further analysis of the data today.  This is not a real-time data source for verification, as discussed before.

See attached AHPS precip. image from this morning. 

Also, attached RADAR summary for that day, and attached MRMS summary.

John

AHPS

RADAR

​

MRMS

​

​

On Mon, Jan 8, 2018 at 4:17 AM, David Barjenbruch - NOAA Federal <david.barjenbruch@noaa.gov> wrote:

Hi,

I was curious if I could get a status update on the discontinuities of the RFCQPE in this thread?  It's still showing up and here's another example from January 7 in the Colorado mountains.  Sorry for the waviness (had to take pictures on phone versus a true screen capture). Interest here surrounds the sharp discontinuities along the RFC boundaries (black lines).

Thanks!
Dave

​​​​

 

On Thu, Apr 28, 2016 at 7:16 PM, Ying Lin <ying.lin@noaa.gov> wrote:

Dave (and John),

    Thank you very much for bringing this up - I see what you're saying, the precip URMA does indeed have discontinuities at the RFC boundaries.  Attached is a 3-pager with 1) 6-hourly URMA ending at 00/06/12/18Z 16 Apr (same as P1 in your pptx), 2) 6h URMA ending at 06Z 16th, and the contributing QPEs from ABRFC/CBRFC/MBRFC (WGRFC is outside of the plot domain), and 3) 24h totals of URMA ending at 12Z 16th, compared to the one from water.weather.gov.  The discontinuity in the 24h total of URMA does indeed jump on this scale.  I spent an inordinate amount of time trying to plot the water.weather.gov 24h Shapefile, but succeeded only in plotting them in monochrome dots.  Anyone knows how to plot a field array Shapefile [not boundary Shapefiles] or NetCDF using GrADS or Python?  Or the RGB color scheme used on the Water site?

    Attached are two files, the OpenOffice 'presentation' (*.odp, created on Linux but should be viewable on Windows with PowerPoint), and the same document exported to PDF. 

    I'll ask Water/AHPS people about how they deal with discontinuities at the RFC boundaries.

Ying

On 04/25/2016 03:32 PM, John Lague - NOAA Federal wrote:

David,
All I can say about the "RFC QPE06" pictured is that it looks different than the QPE we have locally.  I don't know how the data get to you, nor exactly what data it is...so I cannot verify what you are seeing is correct.  The RFC's have been excluded from this process, so it is not really possible for us to answer many detailed questions about it.

You can see from the daily images from water.weather.gov that the 24 hour amounts look fairly good considering we are going over a continental divide.   This web site, and the SRH hourly QPE site are all we really have to go buy.  We don't have any other way to see the neighboring RFC estimates. And there is a good deal of delay before any updates we make get posted.

Much of the issues appear to be differences in the way 24 hour data is time distributed into 6 hour amounts.  This is often problematic, as different gauge networks will be used to provide ratios for time distribution.  In addition, much of the mountain data will be from SNOTEL sites.  SNOTEL data has a great deal of noise is difficult to use even on a six hour basis.

John

On Sun, Apr 24, 2016 at 12:09 PM, David Barjenbruch - NOAA Federal <david.barjenbruch@noaa.gov> wrote:

In an attempt to provide more feedback and improvement in the QPE and verification process, here's another PowerPoint showing the discrepancies in the RFC QPE's for our last significant precipitation event here in Colorado.

Summary of the QPE and where it stands - please let me know if I'm mistaken:

• URMA is stage IV precip remapped to a 2.5 km grid. This is a mosaic from RFC QPE06.
  
• The attached RFC QPE06 graphics in the Powerpoint are those stored at NWS Boulder and available in GFE.  These grids were captured on 04/24/16 for a precipitation event that occurred on 04/16-17/16.
• The attached URMA analyses in the Powerpoint were captured from the mdl.nws.noaa.gov National Blend of models site on 04/24/16 for the same precipitation event.
  
• We realize the URMA analyses are zero for at least the MBRFC here in NWS Boulder, but appears ABRFC is arriving more consistently. CBRFC URMA grids are sporadic.  The new NWSInitsConfig install should ensure the URMA grids are reprocessed. Will this bring in the surrounding grids more reliably, and approximately when can this fix be anticipated?
• There is considerable discrepancies with regard to QPE from the various RFC's, sometimes completely reversing for periods of time (e.g. difference between MBRFC and CBRFC for the 00Z-12Z 17 April time frame, despite a similar flow regime). Are different methods being used at different times within the RFC's?
  
• The water.weather.gov site does not show these variations through time.  From Ying's email the differences should only be slight, but considerable differences are still noted. Those images are attached to the Powerpoint as well.
  

For your reference, there are short descriptions of the Powerpoint in the notes section at the bottom of each slide.  

Thanks again for all your help on this!

Dave

On Wed, Mar 30, 2016 at 5:26 PM, Ying Lin <ying.lin@noaa.gov> wrote:

Hello! I've been away and catching up right now.  Info in Brian's 23 Mar email is correct. Precip URMA is simply the Stage IV 6-hourly mapped to the 2.5km NDFD grid.  To capture as much of the early input and late updates from the RFCs as possible, mosaicking for the 6-hourly fields is remade hourly (if there is input) for the first 24 hours after valid time, then the four 6-hourlies covering a 12Z-12Z 24h period are re-made at 1/3/5/7 days after the ending 12Z  (i.e. at the 12:33Z run).  water.weather.gov doesn't use Stage IV, but there differences should be slight (e.g. they don't include Great Lakes coverage yet).  I understand only from reading this thread that RFCQPE is a Central Region-wide mosaic.

Please let me know if you have other questions. 

Ying

On 03/23/2016 09:21 AM, Brian Miretzky - NOAA Federal wrote:

Jeff and all,

To clarify,

URMA/Stage 4 are essentially the same. Stage 4 is simply remapped to the URMA 2.5 km grid, which could introduce some slight differences. Additional slight differences are possible between water.weather.gov and the Stage 4 NCEP analysis because of slight processing differences between the two. What CR is doing to create the "RFCQPE" I am not sure, but if there are concerns Ying Lin (http://www.emc.ncep.noaa.gov/mmb/ylin/pcpanl/) is your contact for addressing these issues further.

Thanks,

Brian

On Mon, Mar 21, 2016 at 9:52 AM, Jeffrey Craven - NOAA Federal <jeffrey.craven@noaa.gov> wrote:

My recollection from MKX was that the RFCQPE had a latency of about 36 hours.  

JPC

Jeff Craven

Chief, CR Scientific Services (STI)

NWS CRH Kansas City, MO

(816) 268-3110 work

(816) 506-9783 cell/text

http://www.weather.gov/crh/

**NWS Impact Based Decision Support Services:

 The Blend is our friend - our expertise is what matters**

On Mon, Mar 21, 2016 at 8:44 AM, Mike DeWeese - NOAA Federal <mike.deweese@noaa.gov> wrote:

What time is the RFCQPE downloaded?  At NCRFC, we sometimes reprocess the hourly QPE back as far as 72 hours following an event or following a weekend event.  So the most reliable QC'd  data set is not necessarily available on a 24 hour cycle.

Mike

On Sat, Mar 19, 2016 at 8:01 PM, Jerry Wiedenfeld - NOAA Federal <jerry.wiedenfeld@noaa.gov> wrote:

Matt grabs all of the RFC QPE data and mosaics it at region and sends that data via LDM to all WFOs in CR as RFCQPE.

QPF12 and PoP12 verification uses RFCQPE.  1 hour PoP verification is currently using MPE data.  This may change to RTMA but I have found the MPE data is better as long as it is qced at the office.

With the upcoming tech note the urma25 data will be used for verification for PoP12 and QPF12.

Jerry 
On Saturday, March 19, 2016, Jeffrey Craven - NOAA Federal <jeffrey.craven@noaa.gov> wrote:

Thanks Scott.  My understanding is that we grab a file called RFCQPE, which is supposed to be the same as what URMA is using.  

I have included Matt Foster and Jerry since they have set up the verification of our PoP and QPF grids based RFCQPE. 

JPC

Jeff Craven

Chief, CR Scientific Services (STI)

NWS CRH Kansas City, MO

(816) 268-3110 work

(816) 506-9783 cell/text

http://www.weather.gov/crh/

**NWS Impact Based Decision Support Services:

 The Blend is our friend - our expertise is what matters**

On Sat, Mar 19, 2016 at 9:31 AM, Scott Dummer - NOAA Federal <scott.dummer@noaa.gov> wrote:

Paul,

The MBRFC uses DailyQC to help calculate its QPE in the mountainous western portion of it area.  It appears the water.weather.gov image is capturing the DailyQC analysis, while the grids in GFE are using our MPE only grids.  Daily QC relies on point precipitation amounts and extrapolates these into gridded amounts based on PRISM climatological precipitation data which varies based on elevation.

Since this seems like a CRH SSD 'Grids' item, I am redirecting it to John Eise of CRH SSD.  He, or one of his contacts can give a better answer.

John Lague our Senior HAS Forecast might also have something to offer on this as well.

Thanks,

Scott

On Wed, Mar 16, 2016 at 3:14 PM, Paul Wolyn - NOAA Federal <paul.wolyn@noaa.gov> wrote:

Please see this attached powerpoint from Dave Barjenbruch at WFO Boulder concerning RFC QPE discontinuities in GFE.  Is GFE using the same data source at water.weather.gov?

Thanks,

Paul

---------- Forwarded message ----------
From: David Barjenbruch - NOAA Federal <david.barjenbruch@noaa.gov>
Date: Fri, Mar 11, 2016 at 8:27 PM
Subject: RFCQPE & Stage IV Precip
To: Paul Wolyn - NOAA Federal <paul.wolyn@noaa.gov>
Cc: Nezette Rydell - NOAA Federal <nezette.rydell@noaa.gov>, Jennifer Stark - NOAA Federal <jennifer.stark@noaa.gov>, Benjamin Moyer - NOAA Federal <benjamin.moyer@noaa.gov>

Hi Paul,

Thanks for all your input into the RFCQPE, Stage IV precipitation analysis, and URMA.  It sounds like there is some strong movement in that area.  I had a couple more events to share with you about RFCQPE discontinuities.  They are in the attached PowerPoint with a short description in notes.

It appears the water.weather.gov site is using Stage IV Precip, is that correct?  The case examples in the PowerPoint show numerous and wide disparities in the RFCQPE (GFE grids - do you know what "Stage" that is?), but considerably less in the Stage IV Precip analysis.  CRH is using Stage IV precip for verification, correct?

Thanks, and feel free to pass on the PowerPoint if you feel it would be useful as a reference.

Dave B.

P.S. If you'd like me to look at anything in particular for URMA or CRGMAT let me know.  I can try and catch some examples or provide input.

-- 

David Barjenbruch

National Weather Service

Boulder, CO 80305

303-494-3210

--

Jerry Wiedenfeld 

Information Technology Officer 

National Weather Service, Milwaukee/Sullivan, WI

--

Mike DeWeese

Development and Operations Hydrologist

North Central River Forecast Center

mike.deweese@noaa.gov

952-368-2521

 

-- 
Ying Lin
NCEP/EMC/Mesoscale Modeling Branch
NCWCP Cubicle No. 2015
Ying.Lin@noaa.gov

--

David Barjenbruch

National Weather Service

Boulder, CO 80305

303-494-3210

-- 
Ying Lin
NCEP/EMC/Mesoscale Modeling Branch
NCWCP Cubicle No. 2015
Ying.Lin@noaa.gov

--

David Barjenbruch

National Weather Service

Boulder, CO 80305

303-494-3210

--

Paul Wolyn Ph.D.

Science and Operations Officer

National Weather Service, Pueblo, CO

719-948-9429 x766

--

David Barjenbruch

National Weather Service

Boulder, CO 80305

303-494-3210

​

-- 
Ying Lin
NCEP/EMC/Verification, Post-processing and Product Generation Branch
NCWCP Cubicle No. 2015
Ying.Lin@noaa.gov

--

Paul Schlatter
Science and Operations Officer
National Weather Service

Denver/Boulder Weather Forecast Office

303-494-3210 x766

Paul.T.Schlatter@noaa.gov

-- 
Ying Lin
NCEP/EMC/Verification, Post-processing and Product Generation Branch
NCWCP Cubicle No. 2015
Ying.Lin@noaa.gov