RUC testing – Updated 8 June 2006
-
Stan Benjamin, NOAA/ESRL/GSD, Geoff Manikin, NCEP/EMC
-
Other contributors to code changes, testing – ESRL/GSD -
John Brown, Steve Weygandt, Tanya Smirnova
Changes listed below for RUC analysis and model code are in
real-time testing as Feb 2006 - current
-
NCEP/EMC parallel RUC13 cycle (Geoff Manikin has real-time parallel/oper graphics available at http://www.emc.ncep.noaa.gov/mmb/ruc2/para
.)
o NCEP
para RUC13 has been running with the main analysis/model changes since November
2005
-
NOAA/ESRL/GSD development RUC13 cycle since fall 2005 and GSD
backup RUC13 cycle since March 2006.
Overall effects of changes:
¤ Improved
precipitation over oceans and coastal areas.
¤ Improved
RH forecasts
¤ Improved
upper-level winds – related to improved precipitation
¤ Improved
surface fields in coastal areas
Analysis
- More
restrictive observation quality control.
- Now
uses obs-minus-background threshold checks (e.g., ~8 K for temps, 12-20
m/s for winds. More
detailed conditions also set – larger thresholds in stratosphere,
smaller thresholds at surface, etc.). RUC 1h forecasts are used for background.
- Helps
screen isolated problem obs with profiler, RASS temperature, aircraft,
and surface observations, that are not consistently detectable by RUC
buddy-check QC.
- Added
check of mean temperature innovation (O-B) for aircraft reports, flagging
all reports from a given aircraft with at least 2K bias and at least 20
reports in last 1-h period.
(Added 16 Feb 2006 to GSD devRUC13.) – IMPLEMENTED Wed
22 Feb 2006 in operational NCEP RUC13.
- Improved
usage of coastal and ocean surface observations
- Revised
check of observation-model elevation difference for observations over
water grid points in model (e.g., KEYW, KACKÉ)
- Improves
representation of winds and ceiling/visibility over coastal and ocean
areas in RUC analysis.
- Nudging
of soil temperature and moisture (introduced with 13km RUC in June 2005)
is now strongly damped in coastal regions where the surface obs and model
land/water type result in representativeness problems. [This was evident in a 2m temp /
soil temp problem in eastern North Carolina found by EMC in Feb
2006.]
- Matching
of nearby grid points (land or water) to coastal observations for
calculating of innovations (observation-background difference) is
improved. This matching uses
the grid point with the smallest 2m temp innovation, assuming that this
indicates whether the coastal station is under a marine or land influence. With this the overall
mean-absolute difference for METARs is decreased from ~1.4K to ~1.0K,
meaning that the coastal surface obs are now introducing less noise in
coastal areas, especially for temperature.
- Fix to
avoid bullseye in RUC analysis if lowest mandatory level with temp data has
elevation less than station elevation (shouldnŐt happen, but it did for
12z on12 Nov 05 at Birmingham, AL).
- Fix
errors in cloud analysis
- The
current RUC cloud analysis inadvertently removes all rain and snow mixing
ratios under the condition of current precipitation from the nearest
METAR station. This is
exactly the opposite of what should happen.
- Use
of GOES cloud-top data over water – The code for height assignment
for marine stratus to identify the top of the marine PBL in the RUC
background had a bug under a certain condition. With the correction, the RUC initial cloud fields now
depict marine stratus much better than in the current operational
RUC. This leads to
improved coastal cloud forecasts, especially on the West Coast.
- Check
for zero observations to avoid 3DVAR problem in this case with mass/wind
imbalance.
Model
- Fix
to error in Grell-Devenyi convective parameterization over water areas
- G-D
convective parameterization uses an ensemble closure approach to estimate
mean mass flux, but the denominator to calculate this mean mass flux was
in error over water areas (where different closures are used). This problem was inadvertently
introduced with the RUC13 in June 2005.
- The
convective precipitation behavior is more realistic (and generally
stronger) with the fix over water areas (e.g., especially Gulf of Mexico,
but also off West Coast over eastern Pacific, and even over Great
Lakes). Some
differences are also noted over land areas. Upper-level winds affected by latent heating from
oceanic convection are also improved.
- Digital
filter initialization (DFI) redesigned to provide more robust model
behavior and improve initial cloud fields (diabatic DFI)
- The
revised DFI in the RUC model now applies a backward integration first
(rather than second).
- The
forward DFI step is now with full physics, allowing more physical
consistency with latent heat effects and improved initial cloud/hydrometeor
fields affecting 0-2h forecasts of these variables.
- Diabatic
DFI allows the RUC model to run without failure using the
higher-resolution terrain elevation field for model failure cases (19
Nov, 22 Nov) with the RUC model using the original adiabatic DFI with the
high-res terrain elevation.
- The
new diabatic DFI also resets the RH and hydrometeor values back to the
analyzed values after the completion of the DFI. This was found to improve
the accuracy of short-range RH forecasts.
- Higher-resolution
topography (works reliably with diabatic DFI, last bullet)
- Better
orographic effects on near-surface winds (including downslope winds,
blocking) and precipitation.
Post-processing
- Modifications
to ceiling (cloud base) diagnostic
- Change
to eliminate ground fog (lowest 2 RUC native levels) from ceiling
identification to distinguish fog from low ceiling as done with METAR
observations.
- Correction
to ceiling interpolation to 20-km (and 40-km) grid to avoid
waffle-pattern.
- Correction
to equivalent potential temperature calculation
- Correction
to temperature calculation in downscaling of RUC13 data to 5km for RTMA
first-guess. Avoids
problem in situation with very strong inversion near surface to provide
reasonable temperatures along sides of slopes in mountainous areas.
- Correction
to 2m RH calculation so that it is consistent with the 2m T and surface
pressure (at the minimum topography level) such that recalculation of
dewpoint (Td) via 2m RH gives the proper 2m Td (as is done in AWIPS). This change will improve AWIPS
calculations using 2m RH (Td, CAPE, etc.)