Forums

Back

RTMA/URMA Terrain Gradient Issue in Temperature Analysis along Wasatch Front / Great Salt Lake

RSS (Opens New Window)(Opens New Window)
DC
David Church, modified 25 Days ago.

RTMA/URMA Terrain Gradient Issue in Temperature Analysis along Wasatch Front / Great Salt Lake

Youngling Posts: 12 Join Date: 1/8/14 Recent Posts

Lately we've noticed some higher temperature forecasts from the NBM along the benches (transitions between the mountains and valleys), than in the valley floors. I was able to trace this back to the RTMA and URMA analysis which is impacting the bias correction in the NBM. This also seems to happen at some pixels along the boundary of the Great Salt Lake lake shore as well. The issue is fairly ubiquitous at these transition zones in elevation, and does not appear to be limited to any specific observation site issues. The other odd thing is that I actually can't explain how the RMTA and URMA are arriving at the final analysis values. Using the KML files available to see how the obs are impacting the RTMA analysis, the final analysis values for the offending pixels in the KML file are actually reasonable values, but they don't actually match what I'm seeing the gridded RTMA and URMA. Furthermore, it appears the RTMA is too warm at these locations, but the URMA is even warmer at these locations, like the error magnifies between the RTMA and URMA. I've attached some examples are specific points, but the issue is far more widespread than just these pixels. The issue also shows up any day, and seem independent of how warm/cool the day is, we just happened to notice this on the warmer days as this has downstream impacts to the NBM and thus HeatRisk forecasts as well.

Attachments:

MM
Matthew Morris, modified 24 Days ago.

RE: RTMA/URMA Terrain Gradient Issue in Temperature Analysis along Wasatch Front / Great Salt Lake

Youngling Posts: 175 Join Date: 12/6/17 Recent Posts
Hi David,

Thanks for sharing this case with us.  We are looking into it and will follow-up once we know more, likely early next week.

Matt

On Tue, Jun 24, 2025 at 3:21 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

Lately we've noticed some higher temperature forecasts from the NBM along the benches (transitions between the mountains and valleys), than in the valley floors. I was able to trace this back to the RTMA and URMA analysis which is impacting the bias correction in the NBM. This also seems to happen at some pixels along the boundary of the Great Salt Lake lake shore as well. The issue is fairly ubiquitous at these transition zones in elevation, and does not appear to be limited to any specific observation site issues. The other odd thing is that I actually can't explain how the RMTA and URMA are arriving at the final analysis values. Using the KML files available to see how the obs are impacting the RTMA analysis, the final analysis values for the offending pixels in the KML file are actually reasonable values, but they don't actually match what I'm seeing the gridded RTMA and URMA. Furthermore, it appears the RTMA is too warm at these locations, but the URMA is even warmer at these locations, like the error magnifies between the RTMA and URMA. I've attached some examples are specific points, but the issue is far more widespread than just these pixels. The issue also shows up any day, and seem independent of how warm/cool the day is, we just happened to notice this on the warmer days as this has downstream impacts to the NBM and thus HeatRisk forecasts as well.


--
David Church RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/discussions-forums-/-/message_boards/view_message/45549960VLab.Notifications@noaa.gov


--
Matthew Morris
SAIC at NOAA/NWS/NCEP/EMC
5830 University Research Ct., Rm. 2038
College Park, MD 20740
301-683-3758
MM
Matthew Morris, modified 18 Days ago.

RE: RTMA/URMA Terrain Gradient Issue in Temperature Analysis along Wasatch Front / Great Salt Lake

Youngling Posts: 175 Join Date: 12/6/17 Recent Posts
Hi David,

We are still looking into this case study, but we wanted to share our initial findings with you before the holiday weekend.  Please see the attached slide deck for the RTMA/URMA figures valid 20250619/23Z:
First, we'll consider the warmer than expected temperatures along the southern boundary of the Great Salt Lake.  There is a positive analysis increment encompassing the lake in RTMA and URMA, with the magnitude slightly greater in URMA.  This seems to be driven by a MesoWest/UPR station: UP068.  For the time in question, the observation was 92.93F, while the RTMA background was 83.03F, resulting in an innovation of almost 10F.  In URMA, an additional observation from UP472 was used, although it was withheld for cross-validation in the first two outer loops of the assimilation; the observed value was 101.93F, while the URMA background was 89.51F.  These observations appear to be non-representative, and we recommend that they be flagged via the SDM reject list.

In addition, there are some questionable observations along the southern boundary of the lake.  KCC02 reported an observation of 99.23F vs 94.55F.  A colocated CWOP station (AS768) reported for URMA, but not RTMA, with an observed value of 100.13F.  Both of these factors are likely contributing to URMA being slightly warmer than RTMA.  We could try removing the temperature observations from one or both of these stations to see if it improves the analyses.

For the Provo Canyon case, several nearby observations could be contributing to this undesired behavior.  PC034 has observed and background values of 94.73F vs. 88.43F, respectively, while C9635 has values of 98.15F vs 90.41F.  We will be further investigating this case, but suspect that the recursive filter is performing poorly given the terrain in this area.  It may also be worthwhile to explore flagging these stations.

We were unable to locate an analysis increment that corresponds with the circled bench, so it's possible this feature is originating in the background fields.  Please let us know if we are mistaken on this, so we can take another look, if needed.

Finally, for the point case studies you've included, the values that are used to generate the KML file are derived using bilinear interpolation with the 4 nearest grid points, rather than taking the value from the nearest grid point.  In your examples, the points in question are surrounded by cooler grid points, which would tend to reduce the values listed in the KML files.

Please let us know if you have any further questions or would like to proceed with flagging any of the listed stations.

Thanks,
Matt

On Tue, Jun 24, 2025 at 3:21 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

Lately we've noticed some higher temperature forecasts from the NBM along the benches (transitions between the mountains and valleys), than in the valley floors. I was able to trace this back to the RTMA and URMA analysis which is impacting the bias correction in the NBM. This also seems to happen at some pixels along the boundary of the Great Salt Lake lake shore as well. The issue is fairly ubiquitous at these transition zones in elevation, and does not appear to be limited to any specific observation site issues. The other odd thing is that I actually can't explain how the RMTA and URMA are arriving at the final analysis values. Using the KML files available to see how the obs are impacting the RTMA analysis, the final analysis values for the offending pixels in the KML file are actually reasonable values, but they don't actually match what I'm seeing the gridded RTMA and URMA. Furthermore, it appears the RTMA is too warm at these locations, but the URMA is even warmer at these locations, like the error magnifies between the RTMA and URMA. I've attached some examples are specific points, but the issue is far more widespread than just these pixels. The issue also shows up any day, and seem independent of how warm/cool the day is, we just happened to notice this on the warmer days as this has downstream impacts to the NBM and thus HeatRisk forecasts as well.


--
David Church RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/discussions-forums-/-/message_boards/view_message/45549960VLab.Notifications@noaa.gov


--
Matthew Morris
SAIC at NOAA/NWS/NCEP/EMC
5830 University Research Ct., Rm. 2038
College Park, MD 20740
301-683-3758
DC
David Church, modified 8 Days ago.

RE: RTMA/URMA Terrain Gradient Issue in Temperature Analysis along Wasatch Front / Great Salt Lake

Youngling Posts: 12 Join Date: 1/8/14 Recent Posts
Thanks Matt, 
This is very helpful. Sorry for my delay in responding, just catching up after being out on AL last week. 

I think the issues around the Great Salt Lake make the most sense to me, it does seem like some simple black listing of sites could improve the analysis around the lake. The UPR sites seem to often over report temperatures in our experience, so black listing those would be a good start I think. I think at a minimum black listing the sites you mentioned would be a good start, but there are likely other UPR sites that we should do the same with. Would you like me to just work up a list of sites that we could target for a blacklist, and send that over to you?

For the Provo Canyon area obs. I guess my question is, how does the filter deal with elevation variance? Does the filter factor in the fact that the URMA pixel elevation doesn't match the observation elevation? For example, PC034 sits at 5182 feet, but the grid elevation is 6352 feet? The ~1200 foot elevation difference likely accounts for a good portion of the difference between the analysis background and the observation - so I actually don't mind that observation. Where as the CW9635 is sited at 6873 feet, but the background elevation grid appears to be 6437 feet - which casts a little more doubt in my mind as to how good that observation is. I think for Provo Canyon the sharpness of the terrain (even sub-pixel scale) could be causing some of the issues there. 

What I'm still not sure about along the Salt Lake Valley Bench is why there were analysis values that were hotter than any observations in the RTMA and URMA. For instance the MaxT of 102 in the RTMA and 103 in the URMA near 40.7122, -111.8019. There was an observation in that pixel of 98 degrees, and there were surrounding observations as high as 102 in nearby pixels. So at most I could buy 102 (which would be a stretch still), but I really can't figure out where 103 would come from in the analysis. 

Thanks for your help in looking into this more thoroughly. I'll work on trying to come up with a good list of obs that should be blacklisted in the meantime, as it looks like we can probably make some progress with improving the analysis there. 

On Tue, Jul 1, 2025 at 2:42 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Hi David,

We are still looking into this case study, but we wanted to share our initial findings with you before the holiday weekend.  Please see the attached slide deck for the RTMA/URMA figures valid 20250619/23Z:
First, we'll consider the warmer than expected temperatures along the southern boundary of the Great Salt Lake.  There is a positive analysis increment encompassing the lake in RTMA and URMA, with the magnitude slightly greater in URMA.  This seems to be driven by a MesoWest/UPR station: UP068.  For the time in question, the observation was 92.93F, while the RTMA background was 83.03F, resulting in an innovation of almost 10F.  In URMA, an additional observation from UP472 was used, although it was withheld for cross-validation in the first two outer loops of the assimilation; the observed value was 101.93F, while the URMA background was 89.51F.  These observations appear to be non-representative, and we recommend that they be flagged via the SDM reject list.

In addition, there are some questionable observations along the southern boundary of the lake.  KCC02 reported an observation of 99.23F vs 94.55F.  A colocated CWOP station (AS768) reported for URMA, but not RTMA, with an observed value of 100.13F.  Both of these factors are likely contributing to URMA being slightly warmer than RTMA.  We could try removing the temperature observations from one or both of these stations to see if it improves the analyses.

For the Provo Canyon case, several nearby observations could be contributing to this undesired behavior.  PC034 has observed and background values of 94.73F vs. 88.43F, respectively, while C9635 has values of 98.15F vs 90.41F.  We will be further investigating this case, but suspect that the recursive filter is performing poorly given the terrain in this area.  It may also be worthwhile to explore flagging these stations.

We were unable to locate an analysis increment that corresponds with the circled bench, so it's possible this feature is originating in the background fields.  Please let us know if we are mistaken on this, so we can take another look, if needed.

Finally, for the point case studies you've included, the values that are used to generate the KML file are derived using bilinear interpolation with the 4 nearest grid points, rather than taking the value from the nearest grid point.  In your examples, the points in question are surrounded by cooler grid points, which would tend to reduce the values listed in the KML files.

Please let us know if you have any further questions or would like to proceed with flagging any of the listed stations.

Thanks,
Matt

On Tue, Jun 24, 2025 at 3:21 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

Lately we've noticed some higher temperature forecasts from the NBM along the benches (transitions between the mountains and valleys), than in the valley floors. I was able to trace this back to the RTMA and URMA analysis which is impacting the bias correction in the NBM. This also seems to happen at some pixels along the boundary of the Great Salt Lake lake shore as well. The issue is fairly ubiquitous at these transition zones in elevation, and does not appear to be limited to any specific observation site issues. The other odd thing is that I actually can't explain how the RMTA and URMA are arriving at the final analysis values. Using the KML files available to see how the obs are impacting the RTMA analysis, the final analysis values for the offending pixels in the KML file are actually reasonable values, but they don't actually match what I'm seeing the gridded RTMA and URMA. Furthermore, it appears the RTMA is too warm at these locations, but the URMA is even warmer at these locations, like the error magnifies between the RTMA and URMA. I've attached some examples are specific points, but the issue is far more widespread than just these pixels. The issue also shows up any day, and seem independent of how warm/cool the day is, we just happened to notice this on the warmer days as this has downstream impacts to the NBM and thus HeatRisk forecasts as well.


--
David Church RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/discussions-forums-/-/message_boards/view_message/45549960VLab.Notifications@noaa.gov


--
Matthew Morris
SAIC at NOAA/NWS/NCEP/EMC
5830 University Research Ct., Rm. 2038
College Park, MD 20740
301-683-3758

--
Matthew Morris RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/45685352VLab.Notifications@noaa.gov


--

David Church

Science and Operations Officer

ForecastBuilder Program Lead

National Weather Service | NOAA | DOC

Weather Forecast Office | Salt Lake City, UT

801-524-5141 (office)
MM
Matthew Morris, modified 8 Days ago.

RE: RTMA/URMA Terrain Gradient Issue in Temperature Analysis along Wasatch Front / Great Salt Lake

Youngling Posts: 175 Join Date: 12/6/17 Recent Posts
Hi David,

I have added UP068, UP472, KCC02, and AS768 to the reject list for temperature, effective with today's 13Z URMA.  For the remainder of the UPR sites, I have some code that should help quickly identify those sites unless you are able to quickly do so.  But, it might take me some time to get around to this as I'm participating in FFaIR next week.  We will also take a closer look at the Provo Canyon and Salt Lake Valley Bench areas the week of July 21st.  In the meantime, do you want to also try flagging the CW9635 temperature observations?

Thanks,
Matt

On Fri, Jul 11, 2025 at 11:15 AM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Thanks Matt, 
This is very helpful. Sorry for my delay in responding, just catching up after being out on AL last week. 

I think the issues around the Great Salt Lake make the most sense to me, it does seem like some simple black listing of sites could improve the analysis around the lake. The UPR sites seem to often over report temperatures in our experience, so black listing those would be a good start I think. I think at a minimum black listing the sites you mentioned would be a good start, but there are likely other UPR sites that we should do the same with. Would you like me to just work up a list of sites that we could target for a blacklist, and send that over to you?

For the Provo Canyon area obs. I guess my question is, how does the filter deal with elevation variance? Does the filter factor in the fact that the URMA pixel elevation doesn't match the observation elevation? For example, PC034 sits at 5182 feet, but the grid elevation is 6352 feet? The ~1200 foot elevation difference likely accounts for a good portion of the difference between the analysis background and the observation - so I actually don't mind that observation. Where as the CW9635 is sited at 6873 feet, but the background elevation grid appears to be 6437 feet - which casts a little more doubt in my mind as to how good that observation is. I think for Provo Canyon the sharpness of the terrain (even sub-pixel scale) could be causing some of the issues there. 

What I'm still not sure about along the Salt Lake Valley Bench is why there were analysis values that were hotter than any observations in the RTMA and URMA. For instance the MaxT of 102 in the RTMA and 103 in the URMA near 40.7122, -111.8019. There was an observation in that pixel of 98 degrees, and there were surrounding observations as high as 102 in nearby pixels. So at most I could buy 102 (which would be a stretch still), but I really can't figure out where 103 would come from in the analysis. 

Thanks for your help in looking into this more thoroughly. I'll work on trying to come up with a good list of obs that should be blacklisted in the meantime, as it looks like we can probably make some progress with improving the analysis there. 

On Tue, Jul 1, 2025 at 2:42 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Hi David,

We are still looking into this case study, but we wanted to share our initial findings with you before the holiday weekend.  Please see the attached slide deck for the RTMA/URMA figures valid 20250619/23Z:
First, we'll consider the warmer than expected temperatures along the southern boundary of the Great Salt Lake.  There is a positive analysis increment encompassing the lake in RTMA and URMA, with the magnitude slightly greater in URMA.  This seems to be driven by a MesoWest/UPR station: UP068.  For the time in question, the observation was 92.93F, while the RTMA background was 83.03F, resulting in an innovation of almost 10F.  In URMA, an additional observation from UP472 was used, although it was withheld for cross-validation in the first two outer loops of the assimilation; the observed value was 101.93F, while the URMA background was 89.51F.  These observations appear to be non-representative, and we recommend that they be flagged via the SDM reject list.

In addition, there are some questionable observations along the southern boundary of the lake.  KCC02 reported an observation of 99.23F vs 94.55F.  A colocated CWOP station (AS768) reported for URMA, but not RTMA, with an observed value of 100.13F.  Both of these factors are likely contributing to URMA being slightly warmer than RTMA.  We could try removing the temperature observations from one or both of these stations to see if it improves the analyses.

For the Provo Canyon case, several nearby observations could be contributing to this undesired behavior.  PC034 has observed and background values of 94.73F vs. 88.43F, respectively, while C9635 has values of 98.15F vs 90.41F.  We will be further investigating this case, but suspect that the recursive filter is performing poorly given the terrain in this area.  It may also be worthwhile to explore flagging these stations.

We were unable to locate an analysis increment that corresponds with the circled bench, so it's possible this feature is originating in the background fields.  Please let us know if we are mistaken on this, so we can take another look, if needed.

Finally, for the point case studies you've included, the values that are used to generate the KML file are derived using bilinear interpolation with the 4 nearest grid points, rather than taking the value from the nearest grid point.  In your examples, the points in question are surrounded by cooler grid points, which would tend to reduce the values listed in the KML files.

Please let us know if you have any further questions or would like to proceed with flagging any of the listed stations.

Thanks,
Matt

On Tue, Jun 24, 2025 at 3:21 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

Lately we've noticed some higher temperature forecasts from the NBM along the benches (transitions between the mountains and valleys), than in the valley floors. I was able to trace this back to the RTMA and URMA analysis which is impacting the bias correction in the NBM. This also seems to happen at some pixels along the boundary of the Great Salt Lake lake shore as well. The issue is fairly ubiquitous at these transition zones in elevation, and does not appear to be limited to any specific observation site issues. The other odd thing is that I actually can't explain how the RMTA and URMA are arriving at the final analysis values. Using the KML files available to see how the obs are impacting the RTMA analysis, the final analysis values for the offending pixels in the KML file are actually reasonable values, but they don't actually match what I'm seeing the gridded RTMA and URMA. Furthermore, it appears the RTMA is too warm at these locations, but the URMA is even warmer at these locations, like the error magnifies between the RTMA and URMA. I've attached some examples are specific points, but the issue is far more widespread than just these pixels. The issue also shows up any day, and seem independent of how warm/cool the day is, we just happened to notice this on the warmer days as this has downstream impacts to the NBM and thus HeatRisk forecasts as well.


--
David Church RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/discussions-forums-/-/message_boards/view_message/45549960VLab.Notifications@noaa.gov


--
Matthew Morris
SAIC at NOAA/NWS/NCEP/EMC
5830 University Research Ct., Rm. 2038
College Park, MD 20740
301-683-3758

--
Matthew Morris RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/45685352VLab.Notifications@noaa.gov


--

David Church

Science and Operations Officer

ForecastBuilder Program Lead

National Weather Service | NOAA | DOC

Weather Forecast Office | Salt Lake City, UT

801-524-5141 (office)

--
David Church RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/45805326VLab.Notifications@noaa.gov


--
Matthew Morris
SAIC at NOAA/NWS/NCEP/EMC
5830 University Research Ct., Rm. 2038
College Park, MD 20740
301-683-3758
DC
David Church, modified 8 Days ago.

RE: RTMA/URMA Terrain Gradient Issue in Temperature Analysis along Wasatch Front / Great Salt Lake

Youngling Posts: 12 Join Date: 1/8/14 Recent Posts
Thanks Matt! I think we can also go ahead and black list the CW9635 site as well, as the plotting location looks suspicious to me and the temperatures don't appear to match the elevation. 

I also found another bad ob on a bench location near Brigham City that appears to be contributing to the stripe of warmer bench temperatures there: F1767, so let's go ahead and black list that site as well. 

Here is a full list of the active UPR sites in our CWA according to the Synoptic Data API
Station IDs for UPR network in WFO SLC area:
UP005
UP010
UP029
UP030
UP039
UP056
UP066
UP069
UP077
UP094
UP106
UP137
UP142
UP146
UP147
UP153
UP154
UP176
UP199
UP200
UP201
UP218
UP231
UP234
UP242
UP028
UP068
UP138
UP351
UP356
UP456
UP472
UP473
UR296
UR297
UR298
UR299
UR300
UR301
UR302
UR303
UR304
UR305
UR306
UR307
UR308
UR309
UR310
UR311
UR313
UR314
UR316
UR317
UR318
UR319
UR320
UR321
UR322
UR324
UR325
UR326
UR328
UR400
UR406
UR434
UR438
UR466
UR468
UR471
UR539
UR602
UR605
UR606
UR613
UR615
UR630
UR631
UR632
UR633
UR635



On Fri, Jul 11, 2025 at 2:03 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Hi David,

I have added UP068, UP472, KCC02, and AS768 to the reject list for temperature, effective with today's 13Z URMA.  For the remainder of the UPR sites, I have some code that should help quickly identify those sites unless you are able to quickly do so.  But, it might take me some time to get around to this as I'm participating in FFaIR next week.  We will also take a closer look at the Provo Canyon and Salt Lake Valley Bench areas the week of July 21st.  In the meantime, do you want to also try flagging the CW9635 temperature observations?

Thanks,
Matt

On Fri, Jul 11, 2025 at 11:15 AM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Thanks Matt, 
This is very helpful. Sorry for my delay in responding, just catching up after being out on AL last week. 

I think the issues around the Great Salt Lake make the most sense to me, it does seem like some simple black listing of sites could improve the analysis around the lake. The UPR sites seem to often over report temperatures in our experience, so black listing those would be a good start I think. I think at a minimum black listing the sites you mentioned would be a good start, but there are likely other UPR sites that we should do the same with. Would you like me to just work up a list of sites that we could target for a blacklist, and send that over to you?

For the Provo Canyon area obs. I guess my question is, how does the filter deal with elevation variance? Does the filter factor in the fact that the URMA pixel elevation doesn't match the observation elevation? For example, PC034 sits at 5182 feet, but the grid elevation is 6352 feet? The ~1200 foot elevation difference likely accounts for a good portion of the difference between the analysis background and the observation - so I actually don't mind that observation. Where as the CW9635 is sited at 6873 feet, but the background elevation grid appears to be 6437 feet - which casts a little more doubt in my mind as to how good that observation is. I think for Provo Canyon the sharpness of the terrain (even sub-pixel scale) could be causing some of the issues there. 

What I'm still not sure about along the Salt Lake Valley Bench is why there were analysis values that were hotter than any observations in the RTMA and URMA. For instance the MaxT of 102 in the RTMA and 103 in the URMA near 40.7122, -111.8019. There was an observation in that pixel of 98 degrees, and there were surrounding observations as high as 102 in nearby pixels. So at most I could buy 102 (which would be a stretch still), but I really can't figure out where 103 would come from in the analysis. 

Thanks for your help in looking into this more thoroughly. I'll work on trying to come up with a good list of obs that should be blacklisted in the meantime, as it looks like we can probably make some progress with improving the analysis there. 

On Tue, Jul 1, 2025 at 2:42 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Hi David,

We are still looking into this case study, but we wanted to share our initial findings with you before the holiday weekend.  Please see the attached slide deck for the RTMA/URMA figures valid 20250619/23Z:
First, we'll consider the warmer than expected temperatures along the southern boundary of the Great Salt Lake.  There is a positive analysis increment encompassing the lake in RTMA and URMA, with the magnitude slightly greater in URMA.  This seems to be driven by a MesoWest/UPR station: UP068.  For the time in question, the observation was 92.93F, while the RTMA background was 83.03F, resulting in an innovation of almost 10F.  In URMA, an additional observation from UP472 was used, although it was withheld for cross-validation in the first two outer loops of the assimilation; the observed value was 101.93F, while the URMA background was 89.51F.  These observations appear to be non-representative, and we recommend that they be flagged via the SDM reject list.

In addition, there are some questionable observations along the southern boundary of the lake.  KCC02 reported an observation of 99.23F vs 94.55F.  A colocated CWOP station (AS768) reported for URMA, but not RTMA, with an observed value of 100.13F.  Both of these factors are likely contributing to URMA being slightly warmer than RTMA.  We could try removing the temperature observations from one or both of these stations to see if it improves the analyses.

For the Provo Canyon case, several nearby observations could be contributing to this undesired behavior.  PC034 has observed and background values of 94.73F vs. 88.43F, respectively, while C9635 has values of 98.15F vs 90.41F.  We will be further investigating this case, but suspect that the recursive filter is performing poorly given the terrain in this area.  It may also be worthwhile to explore flagging these stations.

We were unable to locate an analysis increment that corresponds with the circled bench, so it's possible this feature is originating in the background fields.  Please let us know if we are mistaken on this, so we can take another look, if needed.

Finally, for the point case studies you've included, the values that are used to generate the KML file are derived using bilinear interpolation with the 4 nearest grid points, rather than taking the value from the nearest grid point.  In your examples, the points in question are surrounded by cooler grid points, which would tend to reduce the values listed in the KML files.

Please let us know if you have any further questions or would like to proceed with flagging any of the listed stations.

Thanks,
Matt

On Tue, Jun 24, 2025 at 3:21 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

Lately we've noticed some higher temperature forecasts from the NBM along the benches (transitions between the mountains and valleys), than in the valley floors. I was able to trace this back to the RTMA and URMA analysis which is impacting the bias correction in the NBM. This also seems to happen at some pixels along the boundary of the Great Salt Lake lake shore as well. The issue is fairly ubiquitous at these transition zones in elevation, and does not appear to be limited to any specific observation site issues. The other odd thing is that I actually can't explain how the RMTA and URMA are arriving at the final analysis values. Using the KML files available to see how the obs are impacting the RTMA analysis, the final analysis values for the offending pixels in the KML file are actually reasonable values, but they don't actually match what I'm seeing the gridded RTMA and URMA. Furthermore, it appears the RTMA is too warm at these locations, but the URMA is even warmer at these locations, like the error magnifies between the RTMA and URMA. I've attached some examples are specific points, but the issue is far more widespread than just these pixels. The issue also shows up any day, and seem independent of how warm/cool the day is, we just happened to notice this on the warmer days as this has downstream impacts to the NBM and thus HeatRisk forecasts as well.


--
David Church RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/discussions-forums-/-/message_boards/view_message/45549960VLab.Notifications@noaa.gov


--
Matthew Morris
SAIC at NOAA/NWS/NCEP/EMC
5830 University Research Ct., Rm. 2038
College Park, MD 20740
301-683-3758

--
Matthew Morris RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/45685352VLab.Notifications@noaa.gov


--

David Church

Science and Operations Officer

ForecastBuilder Program Lead

National Weather Service | NOAA | DOC

Weather Forecast Office | Salt Lake City, UT

801-524-5141 (office)

--
David Church RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/45805326VLab.Notifications@noaa.gov


--
Matthew Morris
SAIC at NOAA/NWS/NCEP/EMC
5830 University Research Ct., Rm. 2038
College Park, MD 20740
301-683-3758

--
Matthew Morris RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/45811343VLab.Notifications@noaa.gov


--

David Church

Science and Operations Officer

ForecastBuilder Program Lead

National Weather Service | NOAA | DOC

Weather Forecast Office | Salt Lake City, UT

801-524-5141 (office)

Attachments:

Attachment

MM
Matthew Morris, modified 1 Day ago.

RE: RTMA/URMA Terrain Gradient Issue in Temperature Analysis along Wasatch Front / Great Salt Lake

Youngling Posts: 175 Join Date: 12/6/17 Recent Posts
Hi David,

The two CWOP stations (C9635 and F1767) and the list of UPR stations that you identified have been added to the reject list for temperature, as of today's 11Z URMA cycle.  Note: UR307, UR321, and UR613 were already on the reject list for temperature following earlier QC requests.  If you continue to observe any issues with the temperature analyses in your area, please let us know so we can investigate further.

Thanks,
Matt

On Fri, Jul 11, 2025 at 4:52 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Thanks Matt! I think we can also go ahead and black list the CW9635 site as well, as the plotting location looks suspicious to me and the temperatures don't appear to match the elevation. 

I also found another bad ob on a bench location near Brigham City that appears to be contributing to the stripe of warmer bench temperatures there: F1767, so let's go ahead and black list that site as well. 

Here is a full list of the active UPR sites in our CWA according to the Synoptic Data API
Station IDs for UPR network in WFO SLC area:
UP005
UP010
UP029
UP030
UP039
UP056
UP066
UP069
UP077
UP094
UP106
UP137
UP142
UP146
UP147
UP153
UP154
UP176
UP199
UP200
UP201
UP218
UP231
UP234
UP242
UP028
UP068
UP138
UP351
UP356
UP456
UP472
UP473
UR296
UR297
UR298
UR299
UR300
UR301
UR302
UR303
UR304
UR305
UR306
UR307
UR308
UR309
UR310
UR311
UR313
UR314
UR316
UR317
UR318
UR319
UR320
UR321
UR322
UR324
UR325
UR326
UR328
UR400
UR406
UR434
UR438
UR466
UR468
UR471
UR539
UR602
UR605
UR606
UR613
UR615
UR630
UR631
UR632
UR633
UR635



On Fri, Jul 11, 2025 at 2:03 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Hi David,

I have added UP068, UP472, KCC02, and AS768 to the reject list for temperature, effective with today's 13Z URMA.  For the remainder of the UPR sites, I have some code that should help quickly identify those sites unless you are able to quickly do so.  But, it might take me some time to get around to this as I'm participating in FFaIR next week.  We will also take a closer look at the Provo Canyon and Salt Lake Valley Bench areas the week of July 21st.  In the meantime, do you want to also try flagging the CW9635 temperature observations?

Thanks,
Matt

On Fri, Jul 11, 2025 at 11:15 AM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Thanks Matt, 
This is very helpful. Sorry for my delay in responding, just catching up after being out on AL last week. 

I think the issues around the Great Salt Lake make the most sense to me, it does seem like some simple black listing of sites could improve the analysis around the lake. The UPR sites seem to often over report temperatures in our experience, so black listing those would be a good start I think. I think at a minimum black listing the sites you mentioned would be a good start, but there are likely other UPR sites that we should do the same with. Would you like me to just work up a list of sites that we could target for a blacklist, and send that over to you?

For the Provo Canyon area obs. I guess my question is, how does the filter deal with elevation variance? Does the filter factor in the fact that the URMA pixel elevation doesn't match the observation elevation? For example, PC034 sits at 5182 feet, but the grid elevation is 6352 feet? The ~1200 foot elevation difference likely accounts for a good portion of the difference between the analysis background and the observation - so I actually don't mind that observation. Where as the CW9635 is sited at 6873 feet, but the background elevation grid appears to be 6437 feet - which casts a little more doubt in my mind as to how good that observation is. I think for Provo Canyon the sharpness of the terrain (even sub-pixel scale) could be causing some of the issues there. 

What I'm still not sure about along the Salt Lake Valley Bench is why there were analysis values that were hotter than any observations in the RTMA and URMA. For instance the MaxT of 102 in the RTMA and 103 in the URMA near 40.7122, -111.8019. There was an observation in that pixel of 98 degrees, and there were surrounding observations as high as 102 in nearby pixels. So at most I could buy 102 (which would be a stretch still), but I really can't figure out where 103 would come from in the analysis. 

Thanks for your help in looking into this more thoroughly. I'll work on trying to come up with a good list of obs that should be blacklisted in the meantime, as it looks like we can probably make some progress with improving the analysis there. 

On Tue, Jul 1, 2025 at 2:42 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Hi David,

We are still looking into this case study, but we wanted to share our initial findings with you before the holiday weekend.  Please see the attached slide deck for the RTMA/URMA figures valid 20250619/23Z:
First, we'll consider the warmer than expected temperatures along the southern boundary of the Great Salt Lake.  There is a positive analysis increment encompassing the lake in RTMA and URMA, with the magnitude slightly greater in URMA.  This seems to be driven by a MesoWest/UPR station: UP068.  For the time in question, the observation was 92.93F, while the RTMA background was 83.03F, resulting in an innovation of almost 10F.  In URMA, an additional observation from UP472 was used, although it was withheld for cross-validation in the first two outer loops of the assimilation; the observed value was 101.93F, while the URMA background was 89.51F.  These observations appear to be non-representative, and we recommend that they be flagged via the SDM reject list.

In addition, there are some questionable observations along the southern boundary of the lake.  KCC02 reported an observation of 99.23F vs 94.55F.  A colocated CWOP station (AS768) reported for URMA, but not RTMA, with an observed value of 100.13F.  Both of these factors are likely contributing to URMA being slightly warmer than RTMA.  We could try removing the temperature observations from one or both of these stations to see if it improves the analyses.

For the Provo Canyon case, several nearby observations could be contributing to this undesired behavior.  PC034 has observed and background values of 94.73F vs. 88.43F, respectively, while C9635 has values of 98.15F vs 90.41F.  We will be further investigating this case, but suspect that the recursive filter is performing poorly given the terrain in this area.  It may also be worthwhile to explore flagging these stations.

We were unable to locate an analysis increment that corresponds with the circled bench, so it's possible this feature is originating in the background fields.  Please let us know if we are mistaken on this, so we can take another look, if needed.

Finally, for the point case studies you've included, the values that are used to generate the KML file are derived using bilinear interpolation with the 4 nearest grid points, rather than taking the value from the nearest grid point.  In your examples, the points in question are surrounded by cooler grid points, which would tend to reduce the values listed in the KML files.

Please let us know if you have any further questions or would like to proceed with flagging any of the listed stations.

Thanks,
Matt

On Tue, Jun 24, 2025 at 3:21 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

Lately we've noticed some higher temperature forecasts from the NBM along the benches (transitions between the mountains and valleys), than in the valley floors. I was able to trace this back to the RTMA and URMA analysis which is impacting the bias correction in the NBM. This also seems to happen at some pixels along the boundary of the Great Salt Lake lake shore as well. The issue is fairly ubiquitous at these transition zones in elevation, and does not appear to be limited to any specific observation site issues. The other odd thing is that I actually can't explain how the RMTA and URMA are arriving at the final analysis values. Using the KML files available to see how the obs are impacting the RTMA analysis, the final analysis values for the offending pixels in the KML file are actually reasonable values, but they don't actually match what I'm seeing the gridded RTMA and URMA. Furthermore, it appears the RTMA is too warm at these locations, but the URMA is even warmer at these locations, like the error magnifies between the RTMA and URMA. I've attached some examples are specific points, but the issue is far more widespread than just these pixels. The issue also shows up any day, and seem independent of how warm/cool the day is, we just happened to notice this on the warmer days as this has downstream impacts to the NBM and thus HeatRisk forecasts as well.


--
David Church RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/discussions-forums-/-/message_boards/view_message/45549960VLab.Notifications@noaa.gov


--
Matthew Morris
SAIC at NOAA/NWS/NCEP/EMC
5830 University Research Ct., Rm. 2038
College Park, MD 20740
301-683-3758

--
Matthew Morris RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/45685352VLab.Notifications@noaa.gov


--

David Church

Science and Operations Officer

ForecastBuilder Program Lead

National Weather Service | NOAA | DOC

Weather Forecast Office | Salt Lake City, UT

801-524-5141 (office)

--
David Church RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/45805326VLab.Notifications@noaa.gov


--
Matthew Morris
SAIC at NOAA/NWS/NCEP/EMC
5830 University Research Ct., Rm. 2038
College Park, MD 20740
301-683-3758

--
Matthew Morris RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/45811343VLab.Notifications@noaa.gov


--

David Church

Science and Operations Officer

ForecastBuilder Program Lead

National Weather Service | NOAA | DOC

Weather Forecast Office | Salt Lake City, UT

801-524-5141 (office)

--
David Church RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/45811950VLab.Notifications@noaa.gov


--
Matthew Morris
SAIC at NOAA/NWS/NCEP/EMC
5830 University Research Ct., Rm. 2038
College Park, MD 20740
301-683-3758