WRF/LSM Implementation and Verification Fei Chen, Mukul Tewari, and Wei Wang (NCAR) John Smart (NOAA/FSL) Collaborators: Ken Mitchell, Mike Ek (NCEP), George Gayno, Jerry Wegiel (AFWA), JinWon Kim (UCLA), QingYun Duan (OH) Sponsored by AFWA and NSF

1
WRF/LSM Implementation and Verification Fei
Chen, Mukul Tewari, and Wei Wang (NCAR)John
Smart (NOAA/FSL)Collaborators Ken Mitchell,
Mike Ek (NCEP), George Gayno, Jerry Wegiel
(AFWA), JinWon Kim (UCLA), QingYun Duan
(OH)Sponsored by AFWA and NSF

• Current Status of WRF/LSM implementation
• Verification Framework
• Issues and Future work

2
Unique and complex aspects of implementation of
land surface models in WRF

• Selection of land surface models (NOAH/OSU LSM,
  CLM, RUC LSM, )
• Background surface fields (landuse, soil texture)
• Initialization of soil state (soil moisture,
  sea-ice)
• Initialization of vegetation state (fractional
  coverage, leaf area index, albedo)
• Requirements from other physics routine
• Coupling strategy and verification
• Need coordination among several working groups

3
Tasks Completed

• Introduce background fields (SI) 1) 30-second
  global USGS 24-category landuse map 2) 30-second
  global hybrid (30-sec for CONUS and 5-min
  elsewhere) top and bottom soil texture 3) 1-deg
  annual mean air temperature as lower boundary
  temperature 4) NESDIS 0.144-deg monthly 5-year
  climatology green vegetation fraction 5)NESDIS
  0.144-deg monthly 5-year climatology albedo
• Initialize soil moisture, temperature, snow, and
  sea-ice from AVN and Eta (SI, mass and height
  version)
• Implementation of OSULSM (Physics, WRF release
  1.2 Beta)
• Inclusion of OSULSM for idealized WRF cases (now
  available for mass version)

4
Work in Progress

• Implement the FSL LSM (by Tanay Smirnova, FSL)
• Implement the Common Land Surface Model (by
  XinZhong Liang, U. Illinois)
• Pre-release of the unified NOAH/OSU LSM (UNO?) by
  NCEP in Feb. 2002 for internal test at NCEP,
  NCAR, AFWA, and UCLA
• A new version of the unified LSM is expected to
  be released soon

5
Work in Progress (Con.)

• Upgrade the unified NOAH/OSULSM to F90 (the
  current WRF/OSULSM coupler is in F90)
• Include a four-layer sea-ice model
• Introduce a few new variables ( total and liquid
  soil water content, fractional snow coverage)
• A few changes in the interface
• Expect to implement the unified LSM by September
  2002

6
LSMs verification framework

• Idealized case (completed)
• theoretical partition of net radiation into
  latent, sensible, and ground heat fluxes, phase
  of soil temperature, etc.
• land-surface/atmospheric interactions (e.g.,
  simple 2-D simulations of sea-breeze like
  circulations)
• Document and provide data for LSMs test
• Uncoupled test (long-term PILPS type data)
• Coupled test (LSM and PBL classic cases
• Select cases (CASES97 and IHOP02)
• model set up provide initial soil and veg
  conditions, complete verification data
• Long-term statistics from real-time and
  retrospective tests
• Get this in standard WRF verification data set

7
An example CASES97 verification

• Three golden cases
• 29 April homogeneously dry soil, heterogeneous
  landuse (green Winter Wheat vs dormant grass) in
  the Walnut watershed, KS,70x74 km2
• 10 May heterogeneous soil, both WW and grass
  green
• 20 May homogeneously wet soil, green WW and
  grass
• CASES97 Data
• 9 surface stations (near surface weather, surface
  heat and radiation fluxes) located over different
  landuse
• enhanced sounding at 3 sites every 90 minute,
  Beaumont site grass Oxford and White Water
  sites wheat
• Aircraft sounding and heat fluxes along flight
  legs
• WRF 10km, 244x214x35, MRF PBL, OSULSM, Dudhia
  shortwave, RRTM longwave, new Kain-Fritcsh
  cumulus, initialized by EDAS 40-km output

8
00Z 29 Apr 00Z 30 Apr 1997averaged over 4
grass sites red model, green obs
Longwave downward
shortwave downward
?
Sensible heat
Latent heat
9
Profile of Potential T from 1000-600 mb 30 April
1997 at Beaumont (grass) model oobs
1530Z
0030Z
2030Z
1835Z
10
Profile of Mixing Ratio from 1000-600 mb 30
April 1997 at Beaumont (grass)
1530Z
0030Z
2030Z
1835Z
11
00Z 29 Apr 00Z 30 Apr 1997averaged over 3
wheat sites
12
Profile of Potential T from 1000-600 mb 30 April
1997 at Oxford (wheat)
1530Z
0030Z
2030Z
1835Z
13
Profile of Mixing Ratio from 1000-600 mb 30
April 1997 at Oxford (wheat)
1530Z
0030Z
2030Z
1835Z
14
IHOP02 Case 12Z 15 12 Z 16 June 2002
Stage IV Radargauge
WRF
IHOP NCAR Surface stations
15
Obs 18Z 48 hour WRF 12Z, 36 hour
St. 3
Green sfc pressure (mb) Purple mxing ratio
(g/kg) Red rain rate (mm/hr)
St. 2
St. 1
16
Central leg 3 stations southwest to Wichita
St. 4
Green sfc pressure (mb) Purple mxing ratio
(g/kg) Red rain rate (mm/hr)
St. 5
St. 6
17
Eastern Leg 2 Stations southeast to Wichita
Green sfc pressure (mb) Purple mxing ratio
(g/kg) Red rain rate (mm/hr)
St. 7
St. 8
18
Initialization of Soil State

• Ultimate solution combine LSM, data assimilation
  techniques, and remote sensing data
• AFWA AGRMET System
• Use observed rainfall, solar radiation, and
  analyzed wind, T, and Q
• Same background field as WRF
• Simulate long-term evolution of soil and
  vegetation state at global scale (40 km, upgrade
  to 20 km)
• NCEP plans to unify the background fields in
  NLDAS
• High-Resolution Land Data Assimilation System
  (HRLDAS) running at the same grid of WRF, using
  HR analysis of surface weather variables,
  landuse, etc.

19
4-month (1998) HRLDAS soil moisture vs Oklahoma
Mesonet observation
5-cm
25-cm
20
Summary and Future work

• Good news
• Tight surface-PBL link valuable for verifying and
  adjusting LSM and PBL schemes
• When underlying surface conditions are correctly
  specified, the temperature profile is well
  simulated
• Temporal and spatial distribution of WRF
  rainfall, surface pressure, low-level T and Q are
  reasonable (compared to 8 stations for two IHOP02
  cases)
• Not so good news
• The mixing layer structure for moisture is not
  well captured (MRF seems too efficient)
• Cannot take everything off the shelf for case
  study (soil condition, vegetation condition,
  etc.)
• Urgent need for WRF realtime runs
• Use AGRMET, or NLDAS, or HRLDAS to initialize
  soil state
• Use weekly quasi-realtime green vegetation
  fraction

21
Future work

• Complete LSM/PBL verification data sets
• SI tasks
• Initialize from ARGMET or EDAS or HRLDAS
• global fixed max albedo over deep snow, annual
  minimum greenness, realtime weekly greenness,
  surface slope index, annual maximum greenness,
  frozen soil empirical derivation
• Interface issues
• A LANDRIVER separated from current PBLDRIVER to
  treat land and inland water body (lake) possibly
  on different grid configuration
• Albedo and roughness length need to pass into
  radiation and PBL schemes
• Need precipitation type, convective and
  non-convective rainfall rate

22
Future Plan

• Real data tests (case studies, Using IHOP data,
  etc.)
• NCAR, NCEP, AFWA work on unified NOAH/OSU LSM
• Release of WRF/Unified LSM by August 2002

23
Combining two NCEP rainfall analysis Utilizing
0.25 gauge-only daily rainfall as primary product
and use hourly Stage IV rainfall to partition the
former into hourly timestep as input to LSM
24-hr rainfall accumulated from 4-km hourly
NCEP Stage IV product
0.25 degree NCEP gauge-only daily rainfall
24
24-h rainfall ending 12Z 20 June 1998
MM5 Control
Stage-II Obs
Eta Forecast
MM5 Wet soil
25
Surface layer (top 10 cm soil)
volumetric soil moisture 180
x 180 km2
Initial time From coarse resolution of Eta
field
46 days later Heterogeneity was developed in
the 3-km domain
26
OSU LSM in the PSU/NCAR MM5 and WRF(Pan and
Mahrt, 1987 Ek and Mahrt, 1991 Chen and Dudhia,
2001)
Canopy Water Evaporation
Transpiration
Turbulent Heat Flux to/from Snowpack/Soil/Plant
Canopy
Precipitation
Condensation
on vegetation
Deposition/ Sublimation to/from snowpack
Direct Soil Evaporation
Evaporation from Open Water
on bare soil
Runoff
Snowmelt
D Z 10 cm
Soil Heat Flux
Soil Moisture Flux
D Z 30 cm
Internal Soil Moisture Flux
Internal Soil Heat Flux
D Z 60 cm
Interflow
D Z 100 cm
Gravitational Flow