Transcript Operational Use of the Rapid Update Cycle

Operational Use of the
Rapid Update Cycle
Stan Benjamin - NOAA/FSL
[email protected]
http://maps.fsl.noaa.gov - RUC web page
COMAP 99
18 August 1999
RUC/MAPS Purpose
• Provide high-frequency mesoscale analyses
and short-range forecasts for:
–
–
–
–
–
aviation
severe weather forecasting
forecasts for public
other transportation
agriculture
2
What Runs Where
• Rapid Update Cycle (RUC)
– Operational Version at NCEP
• Mesoscale Analysis and Prediction System
(MAPS)
– Experimental Version at NOAA/ERL/FSL
(Essentially the same software.
New capabilities tested first in MAPS at FSL)
The 1-h Version of the RUC
Data cutoff - +20 min, 2nd run at +55 min at 0000, 1200 UTC
NCEP Operational Models
Model
RUC
Eta
Name of Duration
Run
RUC
12 hr
Freq.
Domain
Where Develop.
1 hr
CONUS FSL/U. of Miami
Eta
36/48 hr
6/9 hr
N.Amer.+
NCEP/U. of
Belgrade
Global
Spectral
AVN
72 hr
6 hr
Global
NCEP
Global
Spectral
MRF
240 hr
24 hr
Global
NCEP
Global
NCEP
Coupled
Seasons
Key Personnel -- RUC
Development/Implementation
Stan Benjamin
John Brown
Kevin Brundage
Dezso Devenyi
Georg Grell
Barry Schwartz
Tanya Smirnova
Tracy Lorraine Smith
Tom Schlatter
Analysis/model dev, mgmt
Model dev, parameterizations
NCEP impl., WWW, graphics
3-d VAR development
Model dev., parameterizations
Obs ingest, obs sensitivity studies
Land-sfc processes
Obs ingest/sensitivity, NCEP impl.
Interaction w/ NCEP, FAA
Geoff Manikin
Geoff DiMego
NCEP liaison for RUC
Interaction w/ NCEP/NCO, NWS
Uses of the RUC
• Explicit Use of Short-Range Forecasts
– Aviation Weather Center - airmets, sigmets
– Storm Prediction Center - severe weather watches
– FAA - air traffic management (CTAS), CWSUs...
– National Weather Service Forecast Offices
– Airline Forecasting Offices
– NASA Space Flight Centers
• Monitoring Current Conditions with Hourly Analyses
• Evaluating Trends of Longer-Range Models
NWS Forecast Discussion Use of RUC
- Feb-Jul 1999
80-240
40-79
20-39
10-19
8
http://maps.fsl.noaa.gov
9
Hourly Data for 40 km MAPS/RUC-2
Data Type
~Number
Rawinsonde (inc. special obs)
80
WPDN/NPN profilers
31
- 405 MHz
Boundary layer profilers
15
RASS (WPDN and PBL)
15
VAD winds (WSR-88D)
110-130
Aircraft (ACARS)(V,temp) 700-3000
Surface - land (V,psfc,T,Td) 1500-1700
Buoy
100-200
Yellow items new for RUC-2
Freq. Use
/12h
/ 1h
NCEP and FSL
NCEP and FSL
/
/
/
/
/
/
FSL only
FSL only
**NCEP & FSL
NCEP and FSL
NCEP and FSL
NCEP and FSL
1h
1h
1h
1h
1h
1h
**not used since 1/99
pending QC issues
Hourly Data for 40 km MAPS/RUC-2
(cont).
Data Type
~Number
Freq. Use
GOES precipitable water
GOES high-density cloud
drift winds
(IR, VIS, WV cloud top)
SSM/I precipitable water
Ship reports
Reconnaissance
dropwinsonde
1000-2500
/ 1h
1000-2500
1000-4000
10s
/ 3h NCEP and FSL
/2-6h NCEP only
/ 3h NCEP only
a few
/ variable NCEP only
NCEP and FSL
Yellow items new for RUC-2
Real-time observation counts at http://maps.fsl.noaa.gov for RUC-2
and 40-km MAPS
Advantages of q Coords for
Data Assimilation
Analysis
- adaptive 3-d correlation structures and
analysis increments, esp. near baroclinic
zones
- improved coherence of observations near
fronts for QC
Forecast Model
- reduced vertical flux through coordinate
surfaces, leading to reduced vertical
dispersion -- much of vertical motion
implicit in 2-d horiz. advection
- conservation of potential vorticity
- reduced spin-up problems (Johnson et al.
13
93 MWR)
RUC hybrid-b levels - cross-section
Hybrid-b levels
- solid
q levels (every 6 K)
- dashed
No discontinuities
at q/s transitions
Effect of vertical coordinate
on frontal features
Turbulence diagnostic
at FL200 (20,000 ft)
- calculated from native grid
from both MesoEta and RUC
(matched forecast times)
Sharper frontal resolution
with RUC despite coarser
horizontal resolution and
fewer vertical levels
15
RUC-2 Analysis
• QC - buddy check, removal of VADs w/ possible
bird contamination problems
• 3-part analysis (all using optimal interpolation)
1) univariate precipitable water (PW) analysis - using
satellite PW obs - update mixing ratio field
2) z/u/v 3-d multivariate analysis - update qv based on
height/thickness analysis increment, update
psfc from height increment at sfc, update u/v at
all levels
RUC-2 Analysis, cont.
3) univariate analysis of condensation pressure at all
levels, qv at all levels. Also update u/v near
sfc and psfc with univariate analysis with
smaller correlation lengths
• Pass through soil moisture, cloud mixing ratios,
snow cover/temperature (will alter these fields at
future time)
RUC-2 Analysis, cont.
• Vertical spreading (correlation of forecast
error) based on potential temperature
separation (not pressure separation as w/ other models)
• Analysis in generalized vertical coordinate
(code applicable to pressure, sigma, or eta
analysis) except for adjustment at end to
reference potential temperatures and new psfc
• Background is usually previous 1 hr RUC
forecast
Raob sounding
RUC2 grid sounding
Close fit to observations in RUC2 analysis
19
Raob
RUC after fix
RUC before fix
7 April 99 significant-level fix in RUC-2
20
RUCS 60 km Hourly Surface
Analyses
• Draws fairly closely to data
• Persistence background field (1 hr previous
analysis
– some QC problems
– no consistency with terrain
• MAPS sea-level pressure, (Benjamin &
Miller, 1990 MWR)
• Blending to data-void region from NGM
Surface Analyses/Forecasts
in RUC-2
• integrated with 3-d 40 km 1 hr cycle
• dynamic consistency with model forecast
=> accounts for:
– land/water, mtn circulations, sea/lake breezes,
snow cover, vegetation…
• improved quality control - model forecast
background prevents runaway bullseyes
• forecasts out to 12 hr in addition to hourly
analyses
Surface Analyses/Forecasts in
RUC-2, cont.
• Same fields as in 60 km RUCS, plus all
fields available in 3-d system
RUC-2 sfc files (GRIB)
0.3 MB / output time
all variables from RUCS plus
precip
precip type
stability indices
RUC-2 use of surface data
All winds, sfc pressure obs used
T/Td used if abs (Pstation - Pmodel) < 70 mb
- about 90% west of 105ºW, 99% east of 105ºW
ID
FGZ
TUS
SLC
MFR
OAK
SAN
DRA
GJT
RIW
GEG
GTF
UIL
SLE
BOI
GGW
VBG
Eta-48
0*
60
59*
109
18*
12
42
98
104
4*
26
14*
50
55
29
5*
Eta-29
18
13*
68
48*
15*
5*
29*
105
27
11
4*
9*
15*
21*
13
32
|pmodel - pstn|
RUC-40
10 (FLG)
44
59*
67
25
23
** w/I 5 mb of closest fit
34*
65*
16*
1*
14
11*
22
24*
5*
3*
Divergence - 0900 UTC 20 Jan 98
(blue - conv, green/yellow - div)
RUC2 Surface Analysis
Topographical features more
evident with model background
RUCS 60km surface analysis
Little consistency with nighttime
drainage
25
Divergence - RUC2 Surface Analysis - 0600Z 19 April 96
Consistency with topographical features in model
(land/water roughness length variations in this case)
26
RUC surface temperature forecasts
- verification against all METARs in RUC domain
Excellent analysis fit to surface obs (also wind, Td)
3-h forecast better than 3-h persistence
RMS error
Bias (obs - forecast)
national surface temperature verification
99160-99182
national surface temperature bias verification
99160-99182
4
0
2.5
3-h
pers
2
persistence
anx
3-h
1.5
1
temperature bias ( C)
0
RMS temperature error ( C)
3
00z 03z 06z 09z 12z 15z 18z 21z
cycle
Validation time
3
3-h
pers
2
anx
1
3-h
0
-1
-2
-3
-4
00z 03z 06z 09z 12z 15z 18z 21z
cycle
Validation time
Effect of 6 May 1999 Fix
to surface temperature
diagnosis in RUC2
Improved lapse rates in
extrapolation from RUC2
model terrain to different
terrain file (“minimum
topography”) used for
sfc T/Td diagnosis.
28
RUC-2 Model
• Prognostic variables
– Dynamic - (Bleck and Benjamin, 93 MWR)
• qv, p between levels, u, v
– Moisture - (MM5 cloud microphysics)
• q v, qc, qr, qi, qs, qg, Ni (no. conc. ice particles)
– Turbulence - (Burk-Thompson, US Navy, 89 JAS)
– Soil - temperature, moisture - 6 levels (down to 3 m)
– Snow - water equivalent depth, temperature
(soil/snow/veg model - Smirnova et al., 1997 MWR)
RUC-2 Model, cont.
• Numerics
– Continuity equation
• flux-corrected transport (positive definite)
– Advection of qv, all q (moisture) variables
• Smolarkiewicz (1984) positive definite scheme
– Horizontal grid
• Arakawa C
– Vertical grid
• Non-staggered, generalized vertical coordinate
currently set as isentropic-sigma hybrid
RUC-2 Model, cont.
• Cumulus parameterization
– Grell (Mon.Wea.Rev., 1993)
– simplified (1-cloud) Arakawa-Schubert
– includes effects of downdrafts
• Digital filter initialization (Lynch and
Huang, 93 MWR)
– +/- 40 min adiabatic run before each forecast
32
RUC
Digital Filter
Initialization
40 t forward
40 t backward
- digital filter avg
of model values
Produces much
smoother 1-h fcst
Mean absolute sfc pres tendency each t in successive RUC runs
33
Processes in RUC2/MM5 microphysics
(Reisner, Rasmusssen, Bruinthes, 1998, QJRMS)
34
Processes in RUC2/MM5 microphysics
(Reisner, Rasmusssen, Bruinthes, 1998, QJRMS)
35
RUC2 case study - Quebec/New England ice storm - 9 Jan 1998
500 mb height/vorticity - 9h RUC2 fcst valid 2100 UTC
36
RUC2 9h fcst - Surface temp (image), MSLP (beige isobars)
37
N-S cross-section - temperature (isopleths, int = 2 deg C, solid for > 0)
RH (image), 9h RUC2 forecast
38
Montreal ice storm - 9h RUC2 forecast valid 2100 9 Jan 98.
N-S cross sections of RUC2 microphysics
Water vapor mixing ratio / q
Cloud water mixing ratio
| YUL/Montreal
Rain water mixing ratio
Graupel mixing ratio
39
RUC
Land-surface
Process
Parameterization
(Smirnova et al.
1997, MWR)
Ongoing cycle
of soil moisture,
soil temp, snow
cover/depth/temp)
40
RUC/MAPS cycling of soil/snow fields
- soil temperature, soil moisture
- snow water equivalent, snow temperature
MAPS snow water equivalent depth (cm)
5 Jan 1999 1800 UTC
NESDIS snow cover field
5 Jan 1999 2200 UTC
41
MAPS evolved
soil moisture/temperature
vs. observations
May 1999
Bushland, TX
42
RUC - 2 Output Files
• Isobaric main (25 mb, 212 grid)
– 6 3-d variables (ht, temp, RH, u/v, vv)
– 80 2-d variables (prec, indices, spec. level, …)
– ~7 MB / output time
• Surface fields (212 grid)
– 25 2-d variables (p, T, TD, u/v, 3-h dp, precip,
indices…)
– ~0.3 MB / output time
RUC - 2 Output Files, cont.
• AWIPS-211 isobaric/sfc grids (adding vertical
velocity and hourly analyses and 3h fcsts, 212
grids)
• BUFR hourly soundings - same format as Eta
– ~290 stations
– ~1.5 MB for 12-h fcst, all stations (week of 12/8/97)
• Native – ~10 MB / output time
40 km MAPS versus 32 km Eta
June-July
1999
40 km MAPS versus 32 km Eta
June-July
1999
RUC vs. Eta 12-h fcsts
250mb RMS vector error
250 mb wind vector verification
12
Eta better
10
2
9
8
-1
(ms )
RUC minus Eta vector difference
11 4
0
7
6 -2
RUC better
5
-4
98
105 112 119
126 133 140 147 154
julian day (1999)
Comparable skill, potential for ensembles
161 168 175 182
From 80km grids for both models
RUC uses 24h Eta for
lateral boundary conditions
RUC 1, 3, 6, 12h forecasts valid at same time
(against 0000 and 1200 UTC rawinsonde data)
Better wind and temperature forecasts with use of
more recent asynoptic data
48
RUC 3h precip fcst, ending 00 UTC 25 May 99
Difference, RUC fcst minus Stage IV product
3h obs precip, NCEP Stage IV radar/gauge product
3h period ending 00 UTC 25 May 99
Short-range RUC precip verification
Equitable threat score (period ending 00 UTC 25 May 99)
Threshold (inches)
Fcst period 0.01
0.10
0.25
0.50
3h
6h
12h
0.27
0.31
0.33
0.12
0.21
0.31
0.002
0.035
0.171
0
0
0.100
Some skill in short-range RUC precip
- more in winter
49
- improves with longer fcst period
Directions for Future RUC-2
Improvements
(suggested by precipitation verification)
• Improve cloud/moisture analysis.
– Use of advanced microphysics in RUC-2 means that initial
cloud errors can lead to underforecasting. Work is underway
to add satellite, radar and surface data to forecast cloud
fields.
• Introduce fractional cloudiness into the model
– Allow supersaturation at <100% RH within 40 km grid boxes
• Convective parameterization (Grell, includes effect of
downdrafts)
– Gives reasonable performance but still needs
tuning/improvement
Apr 99 emergency change for RUC2
• Correctly uses raob sig-level temp/dewpoint
data now.
• Previously, missed sig-level T/Td data
(TTBB) and forced in linearly interpolated
structures between mandatory levels.
• Significant improvement in RUC grid
sounding structures and in overall RUC
performance
May 99 post-proc fixes for RUC2
• Bug/consistency fixes for diagnosis of sfc T/Td in RUC2.
(fix to lapse rate range)
– Biases in west US for T/Td reduced, 2 °C  0
– s.d. temps over US from 2.0  1.4 °C
(verification against METAR obs)
• CAPE- searches lowest 300 mb, not 180 mb
• More smoothing of isobaric winds in lower troposphere,
near tropopause
• Use of NESDIS ice field
• Much faster running of RUC - 10 procs for all runs
June 99 fix to veg fraction bug
Vegetation fraction in RUC was erroneously set to zero
due to integer/real problem (only a problem w/ NCEP RUC,
not in FSL MAPS/RUC)
Responsible for warm bias from 2100-0900 UTC increasing
during May. Also resulted in dry bias and too little precip
53
July/August 1999 fixes
26 July - fix to moisture in RUC boundary conditions
from Eta
- Eliminate erroneous precip near RUC boundaries
especially over warm oceans
9-10 August - fix cycling of canopy water
- Avoid drying of soil in RUC, restart
soil moisture/temp, avoid warm/dry bias
54
RUC-2 Weaknesses
• Still some precip spin-up problem, despite cycling of
cloud/precip variables, esp. for light precip/overrunning (1-3 hr
late)
• Fix: Add cloud analysis - 1999 - 1st version, allow for
cloud at RH < 100%
• Too much precip over warm oceans, too little near SE coast in
cold season
• Dec 98 fix package helped some - work underway on fixing
tendencies input to Grell convective parameterization
• Daytime convective precip in summer too widespread
• Upcoming fix on tendencies input to Grell scheme
RUC-2 Weaknesses, cont.
• Convective precip forecasts miss many small areas, underforecast
peak amounts.
– Lower equitable threat score than Eta
– more detailed than Eta
• Too much graupel near 0ºC
• Fix: with 20-km RUC, collaboration with FSL and NCAR
on microphysics fixes
• Diurnal cycle of surface temperature a little too weak
– a little too warm at night
• Dec 98 fix package - sfc flux change, radiation fix, GRIB
precision to allow proper soil moisture cycling
• May 99 fix - improve diagnosis of sfc temp/Td diagnosis -significant reduction in bias
RUC-2 Weaknesses, cont.
• Detailed (noisy?) output compared to other models, especially
vertical velocity
– Detail is probably realistic over terrain
• Analysis near coastlines
– does not account for land/sea contrast
– analysis increments over coast extrapolated over sea
Fix: Account for lower horizontal correlation in analysis
when crossing coastline
Fixed RUC-2 Weaknesses
• Analysis sounding structure
– irregular near ground if only sfc data assimilated
Fix: analysis tuning (Dec 98)
Fix: sig-level bug fix (Apr 99) *****************
• CAPE/CIN
– analysis values previously too high in high CAPE areas
– jump between analysis and 1-h forecasts
Fix: CAPE software (Dec 98)
(May 99 - parcel search now in lowest 300 mb, not 180 mb
layer)
RUC-2 Strengths
• Surface fields, especially surface winds
– sfc files
• analysis and forecast
• small
• standard sfc fields plus precip, stability, precip type
• Topographically induced circulations
– sea/lake breezes (scale too large but they’re there)
– mtn/valley circulations
– differential friction effects
RUC-2 Strengths, cont.
• Precipitation fields
– more detailed than Eta (lower FAR but lower POD)
• Snow accumulation
– explicit, not diagnosed (from MM5 microphysics)
• Precipitation type
– uses explicit hydrometeor mixing ratios/fall rates
• Upper-level features
– hybrid q/s coordinate
– winds, PV, temps, fronts, more coherent vorticity
structures on isobaric surfaces
RUC-2 Strengths, cont.
• Lower tropospheric temp/RH
– good fcst sounding structure (esp. after 4/99 fix)
– hybrid coordinate
• Soil/hydro fields
– soil moisture - cycled in 6-level soil model
– surface runoff, canopy water, dew formation, etc.
• Vertical velocity
– available in RUC-2
– good mtn wave depiction, frontal features
• Hourly analyses
– available much sooner than RUC-1 grids
– (4/99 speed-up in RUC processing at NCEP, 12h fcsts
available 30 min sooner, analyses available 8-10 min sooner)
RUC2 precipitation spin-up
1-15 May 1999
15-day total precip
summed from
different RUC/MAPS
forecast projections
Similar precip areas
even from 0-1h fcsts,
spin-up of 0-30%
62
MesoEta
RUC
Theta
63
Mtn wave comparison - MesoEta vs. RUC2
MesoEta
RUC
U - component
Mtn wave comparison - MesoEta vs. RUC
64
W - vertical velocity
Mtn wave comparison - MesoEta vs. RUC
65
RUC/MAPS Web Resources
• Main RUC/MAPS home page
– http://maps.fsl.noaa.gov
• RUC2 discussion forum
– maps.fsl.noaa.gov/forum/eval
• RUC2 real-time data inventory
– maps.fsl.noaa.gov/final.ruc_data.html
• RUC2 Tech. Proc. Bulletin
– maps.fsl.noaa.gov/ruc2.tpb.html
RUC/MAPS Web Resources,
cont.
• RUC-2 diagnosed variables - explanation of each
– maps.fsl.noaa.gov/vartxt.cgi
• RUC-2 evaluation from Nov 97 - Jan 98 field test
– maps.fsl.noaa.gov/ruc2.evalsum.html
• Experimental 36h fcsts run at FSL
• Parallel cycle w/ latest experiments (e.g., cloud analysis)
• AFDs/SFDs using RUC (http://maps.fsl.noaa.gov/sfd)
– used to monitor NWS use of RUC
68
RUC 3dVAR Development
•
•
•
•
•
•
Uses native q-s coordinate
In grid point space
Control variables - z, y, c, q, q
Uses Purser recursive filters
Constraints - weak geostrophy
Includes analog QC (following Dharssi et al.)
69
3-d VAR vs. Optimal interpolation - 40-day parallel cycle test
(60-km 3-h cycle)
(thin - OI, thick - 3dVAR)
3-dVAR giving closer fit to obs in winds/heights for analysis
and 12h forecasts than OI
3dVAR
3dVAR
OI
3dVAR
3dVAR
OI
OI
70
OI
71
RUC 1-h cloud-top forecasts
with and without
GOES cloud-top assimilation
(clearing and building)
(1200 UTC 14 May 1999)
No GOES cloud assim
72
Control - no cloud analysis
Parallel - with cloud analysis
0.7
0.7
0.5
0.5
0.3
0.3
Julian Date
Correlation coefficient between
- NESDIS cloud-top pressure (obs)
- MAPS forecast (hydrometeor mixing ratio > 10-5 g/kg
Cloud-top verification
(calculated every 3 hours)
- strong improvement from hourly cloud assimilation in 1-h forecasts,
73
smaller but consistent improvement even in 12-h forecasts
20km RUC
topography
- early 2000
See
13-km RUC tests
in 1998 FSL
Annual Report
74
13km RUC - 12h forecast - start 0000 UTC 27 October 1997
Precipitation
Surface winds
75
10
20
25
20
13km RUC - 6h forecast valid 0600 UTC 27 Oct 1997
6-h precipitation (cm), wind speed (m/s) in cross-section
76
The Future of the RUC
• Transfer of current 40km RUC2 to IBM SP-2 - July 1999
– faster, distributed post-processing
• 20 km 1 hr version on IBM SP-2
– Probably in early 2000
– 3-d variational analysis
– Cloud/hydrometeor analysis using satellite, radar,
surface, aircraft combined with explicit cloud fcsts in
RUC-2
• Later, assimilation of new data sets: sat. cloudy/clear
radiances (GOES/POES), hourly precipitation analyses,
WSR-88D radial winds, lightning, GPS precipitable water,
sat water vapor winds
The Future of the RUC, cont.
•
•
•
•
– Improved physical parameterizations, including cloud
microphysics (freezing drizzle), surface physics (frozen
soil, high-resolution soil and surface data sets), and
turbulence physics
Higher resolution versions
– 13-15 km/60 level - early 2001
Applications to air quality, coupled air chemistry?
Extensive NAOS observation sensitivity tests
WRF version of RUC
The Future of the RUC, cont.
• Non-hydrostatic -z model under development
– Generalized vertical coordinate
– Nudging of coordinate surfaces toward “grid generator”
• can be set as smoothed quasi-isentropic hybrid coordinate
– treats sub~20km variations (convective clouds, breaking mountain
waves) w/ quasi-horizontal coordinates
– treats >20km variations w/ -z coordinates
– Collaboration between University of Miami (Rainer Bleck,
Zuwen He), FSL (John Brown, Stan Benjamin), and NCAR
(Bill Skamarock)
– Part of WRF model (Weather Research and Forecast NCAR/FSL/NCEP/CAPS) effort - a generalized vertical
coordinate option.
– WRF-based RUC probably by 2005-6 at 5-8 km scale
– 30-min cycle or finer?
Quasi-isentropic option for WRF non-hydrostatic model
Breaking mountain wave simulation - 2 km horizontal resolution
Sigma-z version
Quasi-isentropic version
Thick - q
Thin coordinate
surfaces
80
Feedback
• Send feedback/questions on RUC
performance to the RUC discussion forum.
• Invite us to workshops.
http://maps.fsl.noaa.gov/forum/eval
303-497-6387
[email protected]