Status of the Regional OSSE for SpaceBased LIDAR Winds July 2002

1
Status of the Regional OSSE for Space-Based LIDAR
Winds July 2002
FSL
ETL
NCAR
2
Collaborators

• NOAA/ETL
• Lidar Simulation/ Data
• Atmospheric propagation
• NOAA/FSL
• Regional Nature Run
• Data assimilation
• Assessment
• NCAR
• Quality assessment of model results
• Validation
• NOAA/NWS/NCEP
• Global model background grids
• Advice

3
Goals of Regional OSSE

• Compare to the NCEP global lidar OSSE. Regional
  OSSEs should benefit from lateral boundary
  conditions obtained from the global model that
  has incorporated lidar data.
• Assess impact of space-borne lidar winds on
  regional (mesoscale) weather forecasts.
  Mesoscale models can resolve local forcing from
  surface conditions.
• Demonstrate advantages of assimilation over
  shorter time intervals than for global OSSE.
• Incorporate high-frequency, high-density
  observations that a global OSSE might ignore.

4
Nature runs are the source for simulated lidar
and conventional observations from a regional
model. These observations are assimilated into
an independent regional model. Global models
provide lateral boundary conditions for the
regional models.
5
Regional OSSE Methodology

• Perform the regional nature run (MM5) using the
  global nature run (ECMWF) for lateral boundary
  conditions.
• Extract a suite of simulated observations from
  the MM5.
• Extract lidar line-of-sight winds from satellite
  scanning positions in 4D space from MM5 including
  representativeness error and instrument error.
• Conduct tests with the mesoscale data
  assimilation system (RUC) with and without the
  lidar data, and with various combinations of
  boundary conditions from the MRF model.
• Evaluate the separate effects of the simulated
  lidar data and boundary conditions on regional
  weather forecasts.

6
Flow of Data between Models
MM5 Simulated Observations
ECMWF Background
MM5 Nature Run
Standard Data
ACARS Data
LIDAR Data
RUC Data Assimilation
RUC Forecast Fields
Validation
Results
MRF boundary initial conditions
7

• FSL MM5 REGIONAL NATURE RUN
• Model Physics
• Schultz microphysics
• Kain-Fritsch cumulus parameterization
• Burk-Thompson planetary boundary layer
• RRTM radiation formulation
• RUC Land Surface Model
• Model Grid and Domain
• 10-km horizontal grid spacing (740 x 520)
• 43 sigma levels (model top at 20 km with level
  spacing 800 m)
• Most of North America (includes RUC domain)
• Model Datasets
• Initial and boundary data taken from ECMWF nature
  run
• No soil moisture was available in ECMWF data
  provided by NCEP
• Soil moisture initialized with climatology
  (related to USGS landuse)
• Hourly MM5 output (235 GB total) using 36
  processors on FSL Jet computer

8
ECMWF Nature Run Relative HumidityInitial
Analysis 21 Feb 1993
9
MM5 Nature Run Relative Humidity10 day Forecast
21 Feb 1993
RUC RH
10
ECMWF Nature Run 5-km Temperature Analysis 21
Feb 1993
11
MM5 Regional Nature RunTemperature 10-day
forecast 21 Feb 1993
12
MM5 - ECMWF Nature Run ComparisonsForecast
areal-averaged MSLP Trajectories
Forecast MSLP trajectories diverge, particularly
for days 9 11. However, this drift is
understandable, as shown next
13
ECMWF MSLP 00Z 22 Feb 93
MM5 MSLP 00Z 22 Feb 93
MM5 ECMWF MSLP
Drift in forecast MSLP is concentrated near the
eastern seaboard and is synoptic-scale The
drift results from phase displacements of the
cyclone center and downstream ridge.
14
There is no significant drift in precipitable
water.Drying occurs in the MM5 relative
humidity field at 850 hPa. This RH drift is due
to warming relative to ECMWF nature run. In
summary neither model stands out as better
than the other.
A period of lt 24 h is sufficient to spin up
cloud hydrometeors to realistic values in the MM5
nature run. Cloud liquid water and ice (both
important for lidar simulations) exhibit
realistic spatial structures (not shown here).
15
Assimilation Timeline
ECMWF Nature Run 5 February - 7 March 1993
MM5 Nature Run 11 - 22 February
MM5 Obs Extraction, RUC Assim Forecasts
Verification
Feb 9 10 11 12 13 14 15 16
17 18 19 20 21 22 23
16

• Generation of a Simulated Observation from the
  MM5 Nature Run involves
• Interpolation from the MM5 grid to the
  appropriate time and location of the measurement
  assuming 2001 distribution of conventional data
  (e.g., VAD, ACARS, Profiler) instead of the year
  for the nature run (1993)
• A forward model to compute the observed quantity
  from the model variables if not explicitly
  carried in the model (e.g., lidar LOS radial
  winds)
• Specification of error characteristics
  appropriate for the obs

17
Satellite Track Info
ACARS Track Info
Std OBS Metadata
Simulated Random Errors in Observations Extractor
Standard OBS Extractor
LIDAR OBS Simulator
ACARS Simulator
Simulated Random Errors
MM5 Nature Run Hourly Output
STD OBS Data Files
ACARS Data Files
LIDAR Data Files
18
REGIONAL LIDAR ASSIMILATION

• Use 40-km version of operational RUC20 with 3h
  assimilation cycle
• Assimilate extracted conventional observations
  plus idealized lidar observations
• Boundary conditions (BC) from NCEP MRF model
• Experimental design
• BC only (no lidar or conventional obs)
• BC conventional obs
• Ideal lidar, no clouds, BC conventional obs
• Ideal lidar in cloudy atmosphere, BC
  conventional obs
• Opaque water clouds
• Opaque water clouds ice clouds

19
RUC 151 X 113 gridpoint 40-km domain and
topography
20

• Cycled variables in RUC assimilation cycle
• u, v wind components
• Virtual potential temperature (?v)
• Water vapor mixing ratio (qv)
• Pressure
• Height (diagnostically determined)
• Cloud, land-surface variables
• Control variables in 3DVAR
• ? - stream function
• ? - velocity potential
• Zunb - unbalanced height
• ?v
• ln (qv )

21
Observation Errors (sO) and their Ratio to
Background Errors (sO / sB) used in RUC Data
Assimilation
22
RUC 550 hPa Relative Humidity48h Forecast - 21
February 1993
MM5 RH
23
Instrument model

• Direct detection (molecular) Doppler lidar
  employing a Fabry Perot interferometer (fringe
  imaging system)
• Low satellite orbit (450 km) as proposed by
  industry
• Lidar characteristics 20-W transmitter at 355
  nm, 1-m telescope, 5-sec averaging
• Simulated cloud effects use forecast ice and
  liquid water mixing ratios. Assume water clouds
  are opaque. For ice clouds, we employ a relation
  of ice water content to backscatter and
  extinction coefficients
• Assume an ideal lidar for this initial study,
  so as to duplicate the strategy of global OSSE
  run at NCEP
• Each circular scan is interrupted by 8 stares
  per 1-sec period, consisting of 5 shots averaged
  over 35 km path
• Dual-look scanning of the same atmospheric volume

24
Ice cloud backscatter and extinction
25
LIDAR FORWARD MODEL GEOMETRY
(schematic only, actual model includes earth
curvature)
Lidar LOS wind components
3DVAR analyzes horizontal wind
Neglect vertical velocity in vr
26
Nadir 30º / 930440303
27
Nadir 30º / 930441448
28
8-POINT SCAN Nadir Angle 30 Azimuth Angles
55.05, 117.67, 17.83, 154.88, -25.12,
-162.17, -62.34, -124.95
0
1
-0.5
2
-1
5
3
7
4
-1.5
3
8
Latitude degrees
1
6
-2
4
2
5
-2.5
6
-3
7
8
-3.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
Longitude degrees
29
0
-2
-4
-6
-8
Latitude degrees
-10
-12
-14
-16
-18
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
Longitude degrees
30
Remaining Tasks

• Continuous assimilation and forecasts for test
  period - compare RUC to MM5 nature run
• Comparison of LIDAR, no-LIDAR runs with control
  run, including precipitation impacts
• Final Report due 30 September 2002