Hi Kelly,
Thank you for bringing this issue to our attention. First of
all, a reminder that the minRH in URMA is computed by taking the
minimum RH from the hourly 18Z-06Z analyses. However, the control
variable in the URMA hourly analyses is specific humidity (q) rather
than RH. Also, recall that we generate KML files for URMA that can
be found
here. You can view these in Google Earth, but
restricted observations are not included in these files [complete
files can be shared upon request]. There are no KML files for RH,
as the field is not analyzed directly, while the files for dewpoint
are incomplete, as a station must report T/Q/P in order to be
included in these. Therefore, for moisture issues, we recommend
viewing the files for specific humidity (i.e., "moist").
We started our investigation by taking a look at the minRH for
the May 10, 2025 case, specifically the 22Z URMA (3:00 PM PDT) as a
starting point. Our findings for this particular cycle suggest that
non-representative observations are at least partially responsible
for the higher-than-expected minRH values you are seeing in your
area. The following slide deck shows the specific humidity
background, analysis, and analysis increment figures, as well as
OB/GES/ANL values for select stations.
1) For KEAT, there is a collocated WSAgMet station (WS183) that
is consistently more moist than the background fields. At 22Z on
May 10th, the observed value was 7.06 g/kg, while the background was
4.26 g/kg. There are also several observations on the opposite side
of the Columbia River near Wenatchee that may be worth flagging.
For example, WS058 had an observed value of 6.13 g/kg vs. the
background value of 4.50 g/kg, which appears to be causing a
positive analysis increment in the vicinity.
2) For KGEG, it appears that there may be some influence from
observations to the northwest near Ford/Suncrest. These include
09023, F7716, 07452, F2380, and E1164. There is also a
malfunctioning and/or non-representative CWOP station just to the
northeast of KGEG: G3968. Details are included in the slide
deck. We recommend flagging these stations to see if the minRH
analyses improve in the area.
3) Considering KMWH, there are several nearby mesonet
(SYNOPTIC/DAVIS, WSAgMet, AWS) observations that appear to be
non-representative. These include 03559, 03560, 08744, 0897V,
9042D, WS024, WS181. We recommend flagging these stations, as well.
4) For Pullman, WA (KPUW), it appears that the culprit might be
observations located ~30 km north of Pullman near Garfield. F9216,
02507, 02470, and E7584 are all more moist than the background with
the analysis increment extending south towards Pullman. Note: E7584
was rejected through the prepbufr file for this particular cycle.
We also recommend flagging these stations.
5) For the Stehekin RAWS (STRW1), the figures in the slide deck
suggest that the moist bias is originating in the background fields,
which for temperature, moisture, and pressure is a blend of the HRRR
and NAM Nest data. At 22Z, the specific humidity observation was
3.49 g/kg while the background value was 5.78 g/kg, so it's not
immediately obvious whether these observations are representative.
It's also worth noting that this particular RAWS observation is
being rejected through the diurnal (day) reject list for q.
However, there is a known bug in the GSI code that computes the sun
angles to apply these diurnal lists, so the observations are likely
rejected at all hours as a result. It would require a system
implementation to remove this station from the reject list.
However, if you are able to confirm this is a good quality station,
we can consider removing it later on, should we have the opportunity
for an upgrade.
I hope this helps. Please let us know if you'd like to try
flagging any of the observations listed above or if you have any
further questions. It may take a few iterations to identify all of
the problematic observations, but we hope this is at least a good
start towards resolving this issue. I've started taking a look at
one of the other cases (May 7th) and will follow-up later today or
tomorrow with my findings.
Thanks,
Matt
To whom it may concern,
At OTX, we've observed a consistent wet bias in the URMA
observation grid, particularly with minimum relative humidity
(minRH) values often appearing 5% to 15% too high when copied
into the forecast grid.
This inaccuracy significantly impacts our fire weather forecasts
(FWF) and spot forecasts, as trends are populated from these
observations. We frequently have to make manual edits to correct
errors in the trends, some of which show a trend in the opposite
direction than reality. These trends are read over the radio
during fire season and significantly affect our fire partners.
According to our iMETs, referencing S290/S390 courses, RAWS/ASOS
sites being off by 5% or more is considered unacceptable. We are
hoping for a resolution to this issue or guidance on any
adjustments we might be missing locally.
We have compiled examples from April 27th through May 10th,
including tables and images, which detail these discrepancies.
Please refer to the attached Google Slides for more information.
We are also happy to provide further examples upon request.
https://docs.google.com/presentation/d/10U7RblAe3VPOIcuyEvwhuv9__-hgkASaztSSollPNCs/edit?slide=id.g33d6df8f36c_0_61#slide=id.g33d6df8f36c_0_61
Thanks for any help you can provide!
--
Kelly Butler RTMA/URMA Discussion Group Virtual Lab
Forum https://vlab.noaa.gov/web/715073/discussions-forums-/-/message_boards/view_message/45705351VLab.Notifications@noaa.gov