An FSLRUCRR Proposal for AoR

1
An FSL-RUC/RR Proposal for AoR
Stan Benjamin Dezso Devenyi Steve Weygandt John
M. Brown NOAA / FSL Help from NOHRSC/NWS
Chanhassen, MN - Tom Carroll, Don Cline, Greg
Fall

USWRP AoR Workshop 29 June 2004
2

• Outline of proposal
• Combined approach
• Step 1. Full model-based 1-h (or less)
  assimilation cycle at coarser resolution (e.g.,
  20km (current RUC) ? 13km RUC ? 10km RR)
• Step 2. Non-model downscaling using 1-2km
  topography, land-use, roughness length,
  land/water (e.g., NOHRSC 1km snow
  analysis)
• Step 3. Analysis w/ high-resolution observations
  Mesonet/METAR inc. cloud/vis.., radar,
  satellite

RUC analysis
1-2km downscaled grids
background
1-2km analysis
3

• Advantages for FSL combined-approach AoR proposal
• Extension of existing and planned NCEP
  operational products
• Much less expensive for computer power than
  full-model-downscaling
• Can produce hourly AoRs within 30 min of valid
  time
• Builds on ongoing work to assimilate full
  METAR/sfc obs
• incl. ceiling, cloud levels, visibility, current
  wx (dev RUC)
• to be added to GSI for future Rapid Refresh and
  other NCEP models
• Builds on current hourly 1km CONUS downscaling
  from National Operational Hydrologic Remote
  Sensing Center (NOHRSC). Other downscaling
  methods (e.g., PRISM) also applicable.
• Builds on collaborative GSI development with
  NCEP
• Applicable to Eta/WRF-North American input as
  well as RUC/WRF-Rapid Refresh (use ensemble
  approach).

4

• Outline of proposal
• Combined approach (sequential 3 steps)
• Step 1. Full model-based 1-h (or less)
  assimilation cycle at coarser resolution (e.g.,
  20km (current RUC) ? 13km RUC ? 10km RR)
• Step 2. Non-model downscaling using 1-2km
  topography, land-use, roughness length,
  land/water (e.g., NOHRSC 1km snow
  analysis)
• Step 3. Analysis w/ high-resolution observations
  Mesonet/METAR inc. cloud/vis.., radar,
  satellite

RUC analysis
1-2km downscaled grids
1-2km analysis
5
The 1-h Version of the Rapid Update Cycle at
NCEP
NCEP model hierarchy RUC (1h frequency) ? Eta
(6h) ? Global (6h)
6
10km RUC 9-h forecast surface wind-speed and
barbs overlaid on sfc reports - valid 15z 28 Mar
02
Forecast max wind-speed 48 kts
7
Verify RUC sfc fcsts against all U.S. sfc obs
10-m wind speed
Persist
SUM (AprSep) WIN (OctDec)
2-m temperature
0-h 1-h 3-h 6-h 9-h
Persist
RUC improves surface wind, temp skill down to
1-h fcst Much better than 1-h, 3-h persistence
forecasts
12-h
Fcst Length
8
PBL-based METAR assimilation Use METAR data
through PBL depth from 1h fcst
RUC oper analysis 18z 3 Apr 02 IAD
Effect of PBL-based METAR assimilation
x
x
x
x
9
Assimilation of surface cloud, visibility,
current weather observations into RUC
Goal Modify hydrometeor, RH fields to 1)
force near match to current ceiling/vis obs when
passed through ceiling/vis translation
algorithms 2) improve short-range
predictions Running in real-time test since
Oct 2003 Clearing/building of RUC 3-d
hydrometeor fields Use QC with GOES and
radar Part of RUC cloud/precip analysis w/
GOES, radar, surface obs, background 1-h
forecast
10
Cloud ceiling (m) RUC with and without
METAR cloud assimilation
Oper RUC - w/o METAR cloud assim
Diagnosed ceiling from RUC hydrometeors
18z Obs 17 Nov 2003
METAR Flight Rules
With METAR cloud assim
Corresponding Ceiling height - meters
11
Cloud water mixing ratio (qc), ?
17z 27 Jan 04 analysis After assimilation
of METAR cloud obs
Background 1h fcst
12

• Added assimilation of visibility obs - Feb 2004
• Use FG or BR reports from METARS
• Only when
• Precip is not also reported
• T-Td lt 1K
• Build at lowest 2 levels in RUC (5 m, 20 m)

13

• Characteristics of RUC analysis appropriate for
  AoR
• Hourly mesoscale analysis (digital filter
  essential)
• Designed to fit observations (within expected
  error)
• (incl. Sfc 2m temp (as ?), dewpoint, altimeter,
  wind )
• Consistent with full-physics 1-h forecast
• (most important in physics PBL, land-surface)
• (real-time testing at FSL in RUC20 and
  RUC13)
• Accounting for local PBL depth in assimilation
  of surface data
• Accounting of land-water contrast
• Assimilation of METAR cloud, vis, current wx
• Assimilation of full mesonet obs
• Assimilation of GPS PW, PBL profiler
• QC criteria for mesonet different than METARs
• Assimilation of hourly radar reflectivity/lightn
  ing and GOES cloud-top data into initial fields
  of 3-d hydrometeors (5 types)

14

• Outline of proposal
• Combined approach
• Step 1. Full model-based 1-h (or less)
  assimilation cycle at coarser resolution (e.g.,
  20km (current RUC) ? 13km RUC ? 10km RR)
• Step 2. Non-model downscaling using 1-2km
  topography, land-use, roughness length,
  land/water (e.g., NOHRSC 1km snow
  analysis)
• Step 3. Analysis w/ high-resolution observations
  Mesonet/METAR inc. cloud/vis.., radar,
  satellite

RUC analysis
1-2km downscaled grids
1-2km analysis
15
National Snow Summary Weak upper-level ridging
over the West with weak surface lows continues to
bring warm but unsettled weather. Heat-of-the-day
scattered showers and thunderstorms continue
across the South. more ...

16
NOHRSC Daily Snow Analysis
National Operational Hydrologic Remote Sensing
Center Chanhassen, Minnesota
http//www.nohrsc.nws.gov
17
NOHRSC Hourly analyses at 1 km 1000z 29 June
2004 2m temp, RH - RUC Snow precip, non-snow
precip RUC (later corrected w/ obs) Surface
wind - RUC Solar radiation - GOES
Contour interval 5K
18
NOHRSC Hourly analyses at 1 km 0600z 29 June
2004 2m temp, RH - RUC Snow precip, non-snow
precip - RUC Surface wind - RUC Solar
radiation - GOES
19
NOHRSC SNODAS Snow Model
20
Preprocessing Forcing Data (RUC20)
Full-Res (Internet)
21
Downscaling Solar Radiation
GOES Two-Stream Solar Radiation 0.5 degree Direct
Beam and Diffuse Irradiance
22
Preprocessing Update Data
23
NOHRSC data- Boulder Snowstorm in Colorado
(18-19 March 2003)
RUC forcing
Observed
6 days
24
Enhancements needed for NOHRSC-like downscaling
Wind speed downscaling -- Use u from model grid
scale to calculate wind speed at 1km grid scale
using 1km roughness length Other improved
downscaling? - PRISM - simple PBL/near-sfc
wind models -
Zo at 20km based on USGS 1km data
25

• Outline of proposal
• Combined approach
• Step 1. Full model-based 1-h (or less)
  assimilation cycle at coarser resolution (e.g.,
  20km (current RUC) ? 13km RUC ? 10km RR)
• Step 2. Non-model downscaling using 1-2km
  topography, land-use, roughness length,
  land/water (e.g., NOHRSC 1km snow
  analysis)
• Step 3. Analysis w/ high-resolution observations
  Mesonet/METAR inc. cloud/vis.., radar,
  satellite

RUC analysis
1-2km downscaled grids
1-2km analysis
26

• STEP 3 1-2 km analysis w/ high-resolution
  observations
• Background 1-2km downscaled grids (from step
  2).
• (Step 2 grids are downscaled from Step 1 grids)
• Possible tools all fast analysis steps on 1-2km
  scale
• Barnes- or Bratseth-type analysis using
  innovations (high-res obs minus results of
  downscaling in step 2)
• Simple, fast
• 2dVAR or 3dVAR of innovations, using wavelet or
  digital-filter modeled covariances. (Problem is
  mathematically better conditioned than standard
  3dVAR, also parallelizable.)
• Optimum interpolation (OI)
• Fast, reliable, easy to parallelize
• Ensemble Kalman filters
• also applicable in 3-step method proposed here
• RUC-like use of PBL height, cloud/radar/vis/curre
  nt wx
• Note RUC/GSI 3dvar also assimilate radial winds

27

• Our position 3-d model component necessary for
  AoR.
• But what is the trade-off?
• Only way to allow physical consistency in
  analysis fields for
• topography
• land use (including land-water), land-sfc
  parameterization
• boundary-layer, cloud physics, radiation,
• Essential to produce best possible skill at grid
  points between observations
• Problem with model component for AoR
• Bias in favor of NDFD forecasts that are taken
  from same model as used in AoR.
• Brad Colman (and others) goal AoR should be
  independent as possible from any given model
• Our guarded hopes 1) Steps 2 and 3 will
  provide independence from Step 1. 2) Step 1 can
  have multiple models.

28

• Advantages for FSL combined-approach AoR proposal
• Extension of existing and planned NCEP
  operational products
• Much less expensive for computer power than
  full-model-downscaling
• Can produce hourly AoRs within 30 min of valid
  time
• Build on ongoing work to assimilate full
  METAR/sfc obs
• incl. ceiling, cloud levels, visibility, current
  wx (dev RUC)
• to be added to GSI for future Rapid Refresh and
  other NCEP models
• Build on current hourly 1km CONUS downscaling
  from National Operational Hydrologic Remote
  Sensing Center (NOHRSC). Other downscaling
  methods (e.g., PRISM) also applicable.
• Builds on collaborative GSI development with
  NCEP
• Applicable to Eta/WRF-North American input as
  well as RUC/WRF-Rapid Refresh (use ensemble
  approach).