Forums (Forecast Builder)

Back

RE: Smart Inits

RSS (Opens New Window)(Opens New Window)
Kenneth Cook, modified 8 Years ago.

Smart Inits

Youngling Posts: 15 Join Date: 9/24/12 Recent Posts

I realize this is not a direct FB issue, but we upgraded to the new smartinits on Wednesday.  Since then, we have not been matching with the same starting point as the other offices around us especially with PoPs.  Please advise.

Thanks,

Ken

RL
Robb Lawson, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 8 Join Date: 11/6/15 Recent Posts
Kenneth Cook:

I realize this is not a direct FB issue, but we upgraded to the new smartinits on Wednesday.  Since then, we have not been matching with the same starting point as the other offices around us especially with PoPs.  Please advise.

Thanks,

Ken

I wanted to provide some additional background.  The smartinits we upgraded to was 4.1.   The issue with POPS looks to start around the 6th period of each forecast, however days 6 and 7 appear fine. I attached an image just after the forecastBuilder ran, showing how we match up fine with DDC but not with any other office.  So whatever is happening is also happening at DDC.  I saw an earlier post in this forum where another office was having issues that was caused by the ECE or ECS not being ingested or by the ECMWF MOS not coming in.  I had my IT check both and that doesn't seem to be the issue.

Edit:  Also looks like our QPF grids maybe affected.

JM
Jeremy Martin, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 14 Join Date: 1/9/14 Recent Posts

Ken and Rob,

 

Can you grab a screenshot of what is in your CONSMOS ingredients page for PoP. There may be some data that was missing from that run of the blend.

 

-Jeremy

JM
Jeremy Martin, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 14 Join Date: 1/9/14 Recent Posts

What time period did these grids occur?   Want to see what we had going into the blend at that time.

 

-Jeremy

RL
Robb Lawson, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 8 Join Date: 11/6/15 Recent Posts
Today, we noticed the issues starting with period 6 and continuing through day 5.   However, days 6 and 7 looked fine, which was strange. What loaded through forecastbuilder left at least slight pops in areawide 36 hours longer than everyone else had.
JM
Jeremy Martin, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 14 Join Date: 1/9/14 Recent Posts

If you get a chance, are all the datasets like the CMCnh, ADJECE, DGEX all in your GFE database.  Will keep looking into the issue, hopefully we can find the cause.

 

-Jeremy

RL
Robb Lawson, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 8 Join Date: 11/6/15 Recent Posts
Jeremy Martin:

If you get a chance, are all the datasets like the CMCnh, ADJECE, DGEX all in your GFE database.  Will keep looking into the issue, hopefully we can find the cause.

 

-Jeremy

Thanks for the help Mike!  Jeremy, we do have all of those in our database.  The only ones we do not have are the ALLBlends and from what I've heard, those were discontinued a while back.
MU
Mike Umscheid, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 5 Join Date: 10/26/16 Recent Posts

Jeremy, Robb, Ken,

I found the smoking gun. It has to do with the Canadian global model and it's QPF getting into GFE. Specifically, model "CMCnh" and the native "tp" D2D-Type field. The "tp" grid is what is the input to make the CMCnh IFP-Type grids like QPF and PoP. For some reason, our "tp" grid for the Canadian model is a run-total accumulation of precipitation, instead of a 6-hr accumulation. Therefore, once precip starts accumulating early in the forecast valid times, it will carry forward until the model run end, which is 5.5 or 6 days. This results in 80-100% PoP at say Day 4, some 4+ periods after the storm system is long gone. The 80-100 CMCnh PoP grid is ingested into CONSRaw, which gets into CONSAll, which gets into SuperBlend, which is auto-populated by ForecastBuilder.  So we can blame Canada for our Day 3-5 PoP problems ;-)

In all seriousness though, this needs to be rectified ASAP.  I'm not sure if this is a grib ingest problem into GFE for that model or if it's a SmartInit problem. In other words, I'm not sure if the "tp" grid from the CMCnh that gets into GFE has always been a run total accumulation or if it is 6-hr only like all the other models.  Anyway, at least we know where the problem is.

Mike U

AJ
Andy Just, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 89 Join Date: 6/2/15 Recent Posts

Mike,

  Two things I think that will help resolve the problem you are having with the Canadian. First, did you install the DataAddonsManager and setup the Canadian-NH through that?  The DataAddonsManager should rectify the precipitation coming into D2D.  Second, in a tech order coming out this week for CR, part of the install includes a new release of the National SmartInits. This install will ensure your Canadian parameterInfo file, or the file that helps bring data from D2D to GFE, is correct. 

Kenneth Cook, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 15 Join Date: 9/24/12 Recent Posts

Andy,

We are having the issue still this morning, PoPs are not correct.   Please advise. 

SS
Shane Searcy, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 8 Join Date: 9/26/12 Recent Posts

Ken,

I'll assume the SuperBlend PoP for the overnight run was the one that was off (11/22)?  If so, I took a quick glance at your CONSAll page for PoP (00Z cycle 11/22).  Your ingredients page looks pretty close to mine and EAX's page.  The one difference I did see was your PoP ingredients don't contain the CMCreg input.  I doubt this would cause too much of a difference in PoP, but might cause a few percent discrepancy.  Your ingredients plus neighboring pages can be a good place to find differences.

Perhaps this may provide a place to start.

Good luck.

Shane - CRGMAT

AJ
Andy Just, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 89 Join Date: 6/2/15 Recent Posts

All,

  I spoke with Ken via phone. Turns out something odd going on. The 22.00Z CMCnh QPF was run accumulated in GFE, which obviously impacts PoP especially farther out in time.  However, the new 22.12Z CMCnh QPF in GFE is correct representing each 6 hours.  I'm unsure how this can happen, other than if there is something locally configured somewhere related to the CMCnh, such as a cron. But I'm at a loss.

RL
Robb Lawson, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 8 Join Date: 11/6/15 Recent Posts
We had Matt Foster from up at CR poke around our system and he found a script running on one of PXs that looked fishy.  He stopped it, so hopefully that will fix it. Just frustrating we seem to be the only office having these issues.
MU
Mike Umscheid, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 5 Join Date: 10/26/16 Recent Posts
Mike @ DDC here,  I've attached our CONSMOS ingredients page for PoP for 11/04 12z ... since our PoP was identical to ICT.
 
Jeremy Martin:

Ken and Rob,

 

Can you grab a screenshot of what is in your CONSMOS ingredients page for PoP. There may be some data that was missing from that run of the blend.

 

-Jeremy

JM
Jeremy Martin, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 14 Join Date: 1/9/14 Recent Posts

Thanks Mike!   Can you grab the CONSAll ingredients for both Pop and PoP6?  I will compare them to ours. The MOS part looks the same as we have.

 

-Jeremy

MU
Mike Umscheid, modified 8 Years ago.

RE: Smart Inits

Youngling Posts: 5 Join Date: 10/26/16 Recent Posts

Jeremy:  Attached are CONSALL from 12z this morning for PoP and PoP6.  I'll be curious to see if you find anything.  Thanks Jeremy!  (PS: I'm on evening shifts)

Mike U